Consciousness Live! with David Chalmers (S3 E2)

0:00 all right we are live all right joining me today is an ultra special guest david chalmers so Dave for those who don't know you can you say something about who you are and what you do sure my name is David Chalmers I'm professor of philosophy at New York University where I'm also co-director of the Center for mind brain and consciousness and you know I guess I work on many aspects of philosophy but especially the the philosophy of mind and consciousness in the way it connects to the to the

0:32 science of consciousness so I've been coming at consciousness from many different angles for I guess for many years now I'm also yeah really interested a whole bunch of other issues in philosophy language metaphysics epistemology science and also things like the foundations of physics foundations of cognitive science and so on so but I guess consciousness is at the core yeah is it true that – I think I read somewhere that you said that originally you kind of just want to think about consciousness but then you realize that trying to answer those questions sucks

1:04 you into all those other traditional areas of philosophy yeah pretty much um I remember years ago when I was like a math student just reading philosophy reading these philosophers going on about Hesperus and phosphorus and think oh my god that's just boring semantic issues what's this all about but I found that to think about consciousness in the mind-body problem rapidly I was thinking about for example metaphysics that I've got me thinking about language that got me thinking about epistemology that got

1:37 me thinking about foundations of science so before you know you're getting into almost all of philosophy I did swear at one point that I was never gonna write a paper in ethics I think I've still stuck to that so far but I feel myself weakening on that one yeah I think the last section of your book on value maybe is gonna break that rule at some point yeah yeah a very couple of chapters in yeah yeah I mean speaking of which I this book I don't know if people know that you're working on this book I got

2:09 reality 2.0 is that the title of it that's the working title reality 2.0 virtual worlds and the great problems of philosophy I don't know it's very much a working title many people have told me you can't use that it's such a cliche it's like twenty years ago so if anyone's got a better way of saying reality 2.0 without the 2.0 let me know because it does capture what the book is about it's about artificial realities virtual reality the idea of the world is a simulation and so on and using the idea of artificial

2:43 realities to explore the concept of reality and to explore all kinds of philosophical issues whether pisum illogical issues such as skepticism metaphysical issues about the nature of reality God mind value there's meant to be a book that anyone can read but also tries to really get into some substantial philosophy I've seen a sneak preview of it and I think it does that actually I think it's really aimed at the at the right level I think it's gonna be a great book when it comes out we've been waiting for a book

3:15 like this I do notice that I thought I heard you one time and a talk or something say you were trying to use the word techno philosophy like Norrell philosophy but I noticed that wasn't anywhere in the book anymore oh yeah that'll be in the introduction which isn't which isn't written yet at one point I talked about I thought about calling the book techno philosophy and actually uh my literary agent who is you know he was promoting the book to the publishers you can't call a techno philosophy that sounds too academic from that but maybe that'll be in the introduction I do

3:47 think of it I mean this is inspired by Mura philosophy patricia churchland term which has become ii mean have a two-way interaction between philosophy and neuroscience thinking philosophically about neuroscience but also using neuroscience to think about philosophical questions so that's how I think of it with techno philosophy we're using philosophy to think about issues about technology but we're also using these technological tools to gain insight to thinking about some very traditional philosophical questions yeah that's some love pop but I'm trying to

4:21 do in the book yeah that's I think that's really it's time for that to happen so I I mean there isn't anything since I would guess what was that book that guy wrote this book a philosophy of our virtual reality or something I was a while ago I forget who it was actually he's maybe a non-academic philosopher on philosophical issues about virtual reality yeah there's been a few things here and there there's a lot of people right on video games at least connect so people like Mark Silcox and John Cogburn and a bunch of other people who work in this sort of the philosophy of gaming

4:56 from heaven or Jesper Juul and others have written about video games as a as an art form and as a and have talked for example about the ontological status of those I think the issues they have begin to connect to the issues about virtual reality and and about artificial realities more generally but also video games are something of a special case right a fictional they involve role-playing and so on where I think the issues about VR in general or in some ways in some ways broader I mean people have thought about because people think

5:29 about these old ideas of the world is a simulation there's the classic Nick Bostrom paper on that I wrote a paper on this years ago called the matrix as metaphysics which was actually originally for the for the website of the movie The Matrix where they wanted to get some philosophers involved in writing about philosophical issues that arose from the the movie I'd been thinking about writing a technical paper called in that meant as a metaphysical hypothesis about the whole brain-in-a-vat idea so yeah I could totally reframe this around the movie

6:02 and called it instead the matrix as metaphysics that's actually one of my favorite papers I tried to argue that yeah even if we are in a simulation like The Matrix nonetheless all this is perfectly real many people think now this would all be then fictional or an illusion my line is no even if we are in in the matrix all this is real and in a way that kind of idea is that that is that the foundation yeah what am i giving in this book to reality 2.0 them the rough idea is that yeah virtual reality is a kind of genuine reality

6:35 it's not a second-class reality it's not a fake or fictional reality kind of genuine reality right so this is what you called digital realism yeah virtual realism sometimes for the end of things in virtual reality are real and in particular yeah did virtual Digitalism I think I I call it in some places when you're in a virtual reality or a simulation and so on you're interacting with real objects they're just digital objects something like data structures on a computer but they know less they know this real for that right

7:08 so so the basic idea I think is that it as I understand the argument roughly is that here you're an actor excuse me you're interacting with something which is lawfully guiding the behavior of the digital object so there's bits at the computer level which are causally interacting with each other and so that there'd be no real reason to distinguish what's going on at that level from what's going on at the micro physical level and what we call physical reality it's just a way of interacting with an object well they're different but I think in some ways no less real I mean

7:42 the easiest case to start with is the one where it turns out that all along our physical reality has been grounded in some digital reality as in the matrix case as on the simulation hypothesis everything experience is and always has been computer simulation I guess on that some people say oh no well then this would be none of the things we experience would be real this is a skeptical hypothesis I'm gonna say no these things are still real they're just digital I mean just physics even physics exist more or less as we thought it was it's just the physics is realized by an

8:15 underlying digital level which is not how we actually think it is but it doesn't mean that there's no physics there it just means there's a level of digital processes I'm beneath the level of physics and there's something else underneath that now if we're not if we lay out if we set aside the simulation hypothesis and suppose we're not in a simulation then okay our world is presumably non digital unless physics turns out to go in a interesting digital direction anyway but we can then contrast it with say vote for realities

8:50 the kinds we experience when we put on a VR headset and so on it's okay well these ones are digital realities so they're different physical reality will be non digital virtual virtual realities will be virtual so they're different realities in that case but I still don't want to I still want to say in some sense the VR the virtual reality is still real it's not real and it would have been if we'd been living in a virtual reality our whole life it'll be different from the reality we've grown up in so we might reasonably have different attitudes to it because of that we might especially value the reality the original reality

9:24 the one we've been embedded in and grown to value but nonetheless I think virtual realities are can still be perfectly real that could be places where you could live a meaningful life we could transition into virtual reality spend a good part of our lifetimes there and have a perfectly meaningful life there again it will be different in various respects from a non virtual reality but still meaningful still valuable so I guess that's where I start to ver veer into questions about about values that may eventually break my resistance so I

9:59 know you've draw a distinction between a virtual world and a virtual reality is is a virtual world is just one that's not immersive completely whereas a virtual reality is so the world of warcraft is a world but not a reality is that the right way of using those terms yeah that's how I use them in the book and there's also an article I published about the stuff that came out a couple of years ago called the virtual and the real and there I guess I used what's virtual reality it's a fairly standard definition a virtual reality environment is an environment which is immersive

10:33 interactive and computer-generated three conditions immersive basically means you experience it yeah so three dimensionally from the center of a perceptual world perceiving it as if it's a world around you that's the feeling of being there interactive means it's not just you perceive this reality you can affect it to what you do will have effects on that reality and computer-generated means all those processes the inputs you're getting around it in the processes of a computer so ordinary reality is a most evident

11:07 active but presumably it's not computer generated unless at least unless the simulation hypothesis is a is true so yeah so VR and standardly construed has to meet those three conditions then you might ask that what about say video games like you know World of Warcraft or even Second Life played on a on a desktop on a screen it doesn't fully meet the immersiveness condition because it's just on this flat screen it's not it's not done it's not the 3d feeling of being there immersion

11:41 so it's not strictly speaking virtual reality but I think the way people commonly talk about this they nonetheless call these worlds of virtual worlds so following that practice I say well to be a virtual world you don't have to meet the immersiveness condition just being computer-generated and interactive is good enough then we can also entertain hypotheses like what if it's not interactive then it's just like a 360 video very passive maybe we're not gonna maybe it's stretching things to call that VR or even a reality or a world at all but you know there are

12:15 different grades of these things but the core case is when you meet those three conditions immersive interactive computer-generated but you still gonna want to say that the digital objects in it you know a world are still as real as they are in the virtual reality yeah that's right I think I want to make the probably the ontological point about the reality of the objects applies equally inside the mirror virtual world it's not immersive I mean there's interesting questions about some of the other theses I talked about one is that perception inside virtual reality needn't be an

12:49 illusion it may be in some cases when you don't know you're in VR someone puts a headset on you you might perceive all this is going on in the regular physical world when it's going on in the virtual world but I want to argue that at least for a sophisticated user of VR your perception didn't be illusory or hallucinatory you can be having vertical perception of the digital objects around you you perceive them as virtual objects and that's what they are it's an interesting question to what extent that point transfer to the case of a virtual world on a

13:22 desktop I think probably something like it transfers but it gets a bit more complicated because you've now got this whole screen like representation that's rather different from ordinary perception as opposed to the VR case where the experiences really do seem to be kind of continuous with ordinary perception yeah that's what I was kind of gonna ask you about is whether they're whether you would allow some kind of like levels of reality you know some sometimes people talk about these things so maybe the digital objects and world warcraft are real but not quite as realism one in a fully immersive world

13:54 because of the merge Civet ii of it because there's you know tactile sensations you can interact with it more fully but it sounds like what you're saying is that you're sort of gonna resist that that kind of moves well there's degrees of virtuality let's say for sure and there's degrees of immersiveness and in general there's different criteria for reality I'm not sure that I've got a single one-size-fits-all criterion for what it is for something to be real but at one point in the book I go through various different criteria there's um there's you know Barclays dictum to be is to be

14:27 perceived okay well most people don't quite accept that one be my thing okay someone less a heuristic guide if something is perceivable or observable that's that's at least some indication of it being real there's Alexander's dictum to be real as to play a causal role yeah I was gonna just not to interrupt too fast but that you that you're attributing to Australian a philosopher what's it Armstrong called that the le attic principle because of the ancient Greek guy goes back a long way but I think it's for whatever reason

15:02 this it gets cold Alexander's dictum pretty obvious pretty regularly these days okay I don't think it stops with him this is just a general instance of the effect right there whenever you have X's law it's almost never originated by X like someone or others law of political economy and whoever it is Stiglitz is local or actually the names together on law okay so good go on so there's like to be is to be perceived I mean it'd be fair Berkeley was talking to

15:33 Persis and it is hard to see if you're an empiricist what else I guess maybe you know Quan Yin to be is the valuable variable would be more of a modern in Paris's take on that yeah I mean many of us want to allow things which are not perceived may perhaps not even perceivable but then we want to say well I can still be real and then okay then they might be real and virtue of meeting some of the other conditions like being playing a closer role for example another one I like comes from philip k dick where he says that something is

16:07 real if when you stop believing in it it doesn't go away next this is basically a way of putting what you know philosophers called mind dependence mind independence it's basic you know things in your mind don't stop it from existing changing your mind its existence doesn't depend on the state of your mind now of course you don't want to make that a general criterion for reality because then what about states of your mind maybe they depend on my mind but they're still real numberless for an awful lot of things in the external world one of the things that matters when thinking about their reality is a certain kind of

16:39 independence of the mind yeah I was wondering about the philip k dick stuff because he's sort of well known I think for having schizophrenia right and having some notebooks that were filled with all sorts of strange things one tends to get the feeling that maybe he says stuff like that because he's having these kind of maybe delusions and he wants to count them as real it could be I don't know that much about his his biography I do not he wrote an awful lot about alternate worlds dream realities artificial reality slavery realities seems to mean a very major theme in his

17:15 uh in his work right but is that yeah it was a video where he actually comes out and say and like I said a science fiction convention he says I think maybe we really are in a simulation right now back in the 1970s Wow he's a forward thinker but I mean I guess what I was trying to say was is that a problem for the dictum that if you have some kind of you know if you have like a dilute on a cap gras delusion or something you think you're dead or your loved ones have been replaced by by identical look it's that or not them that's not

17:48 going to go away simply if you I mean if it's a really powerful delusion you may even sort of feel that you're deluded and still in the B and the gripped of it so would that end up meaning that those things are real is that kind of a result of that dictum yeah maybe it shows that the philip that the philip k dick version of it is a bit too bit too strong you stop believing in it but there's still some pre belief process like some perceptual process right it makes it real for years but probably yeah one wants to generalize it to something like not just independence of what you believe but independence of

18:21 your mind in general in this case it looks like well maybe doesn't depend on your beliefs but it depends on your perceptual States right and so on so then is still mind dependent and at least to that extent not fully real but i think of what dick is trying to get at is at least one version of this mind independence criteria yeah i mean to go back to some of the other things i guess i was a little bit surprised at the way you treated berkeley and Descartes in your analysis mm-hmm I thought that at one point you were you ended up saying that even if we were in an evil genius scenario that we would still have our

18:55 ordinary beliefs preserved and things would still be real there still be a table in front of me and my coffee is sitting on and stuff is is that your that they not not only in the so that's not a virtual reality since that's not computer generated so it fails on that third part of the criteria its immersive and all that other stuff but is that is that right that you're gonna count those things as real if this is a which I policies we are now is it the god the evil genius hypothesis evil genius hypothesis yeah I mean it depends on exactly how the evil genius is doing it but if the evil genius is doing it by

19:27 roughly having a model a very detailed model of the world in its head and using that to simulate all the to feed experiences to you based on the simulation then I think the evil genius is basically playing the role of the computer in the matrix and what we're perceiving is real although it is processes in the Evil Geniuses head so it's to that extent mind dependent not to paint it on your own mind but depends on someone else's mind insofar as you take that is your criterion for reality you might say as to that extent less real but otherwise I think it's

20:01 more or less analogous to something like the simulation case and also Barclays case where it's God is I take it more or less analogous to this to God's mind he's in this case playing the role of the computer that's something independent of your own mind that's generating the experiences in your mind so I would count that here as yet relevantly analogous and real if we're in the evil demon world then it turns out that yeah that the things I'm interacting with here are genuine but grounded in the demons mind and same for Barclays God hypothesis okay so yeah

20:36 because I thought so I guess I misunderstood part of what you were saying there because I thought what you ended up saying was that it's not that way for Berkeley but it is that way for Descartes and I guess you know I must have it say about Bach I may have forgotten my own views here yeah I mean I'll have the thing pulled up right in front me but I thought that you were saying that that Barclays view doesn't count as a simulation view I must be wrong a habube is good yeah um I think God is kind of playing the role of the

21:11 least one way of thinking about God as God is playing the role of the simulator another way I think about God is God doesn't need to do any of that right – simulation is head and just automatically generates this stuff and then I don't know don't know what to say about it oh I see so you're building quite a bit into the fact that God's doing having a model in mind in order for yes oh yeah I guess I'm thinking I thought Berkeley not that I have any real expertise on Berkeley I thought that Berkeley thought that when you have an idea of the table this is somehow reflecting or picking up on God's idea of the table so God has all these ideas which are

21:44 somehow interacting and as the world develops right yeah I think you're bringing evolution among God's ideas if that's the case then it looks like you know God's big set of ideas is relevantly analogous to a computer running a simulation okay it's okay also different in many ways in that limited way then you would accept SAS per kit Bay right I mean that in in the Berkeley world that is what real is yeah well I guess if you think about I mean there's a there's a kind of a really basic like

22:18 first lecture on box me wait to give up Berkeley where it's all stuff in your own mind right anything's real if it's perceived by me and then the god picture I think basically ends up giving up on that because you know there could be things that are real like the tree in the quad never never perceived by me but as long as it's perceived by God so to be real now as to be perceived by by somebody including God I take God's perceptions and really the ones that wear the rubber the rubber hits the road so I guess I

22:54 don't agree with Berkeley that to be is to be perceived even to be perceived by God but I agree with a more limited thesis that to be perceived by God in this way in the context of a model fed to us suffice us to be real okay so yeah interesting so then how is it that the simulation hypothesis is different from idealism more generally if you allow that God is playing this role and that these you know external things are really ideas and gods are perceptions in a god-like mind on your view they'd be you know digital processes in a computer

23:28 but so maybe the substrates gonna be different but ultimately isn't it sort of similar well the first thing to say is idealism requires a mind at the base Berkeley has a mind at the base but suddenly these general these general realism about virtual reality that I have in mind doesn't require anything like a mind at the base of mind is just one version it could be it could equally be a computer involving no mind at all well could it really though I mean the idea that's really driving this for me

24:04 is not idealism so much as structuralism what matters for a conception of the external world is a certain kind of structure especially causal structure which could be realized in a man ordinary physical world could be realized in a computer or it could be realized in God's mind can be realized in lots of little bits of justice as on the penstock hypothesis so there's a whole different for different versions of this it's so the so if you take the computational version of it

24:37 these fundamental computational processes aren't gonna have anything in common with minds yeah well anything has a lot of things in common with anything so isn't there at least I'll have their causal role in common and a certain kind of structure in common but if we're not pensar castes and they won't literally be minds there'll be non mental bits of fundamental reality interacting with each other yeah but what makes them not mental I guess is what I'm asking oh well I suppose if you were an extreme functionalist you'd say that they'd be

25:12 non mental mental because of some role that they play it would be a very very extreme functionalist it's like a functionalist pensar caste but otherwise what's the basic reason for wanting to call these things mental any more than I want to call the processes inside my desktop computer mental well I'm wondering why you don't want to call them mental what motivation for calling the mental well they represent states of affairs in the world and my sort of their be kind of you know counted as a sort of awareness of those states if you

25:45 push the notion awareness or now I'm wondering why you're not gonna say something I'm not endorsing it or saying you have to UM just I do think I wonder what like at that point what's really the distinguishing factor unless you were going to bring consciousness in or from your normal consciousness even the claim that they represent as a substantive one I mean it some digital processes represent things in the world some don't in my view they're just digital processes interacting with other digital processes so in my view not all computation involves representation if you look at most of the stuff going on

26:17 inside my my iPhone or my my Mac I'm you know some bits of it are engaged in representation but a lot of bits or not even for those bits which are I guess I wouldn't be especially inclined to call the mental any more than I want to call like the woods on a words on a page of a book mental that's representation but it's not mental representation it's well in the book its linguistic representation in a pictures pictorial representation in a computer it's digital representation those are all I

26:50 guess those are all varieties of representation but I find to distinguish all of those from mental representation well the first two I guess I would want to say they're derivative on mental representations by obviously you don't think computation I know there's this general line out there coming from Searle that all computations are observer-relative and you have a line against that which I think it's pretty good actually but I also individuate computation independently of representation I think computational structure algorithmic structure is a matter of causal structure and close interaction all this is prior to the level of

27:23 semantics representation now as a matter of fact I think yep I think these do like as many ways in which computer processors can represent things in the world some of those I think are observer dependent some of them come from us as a user but some of them some of them may not they may be due to causal to causal tracking to certain kinds of isomorphism and so on I don't want to deny that there can be there can be genuine representation you know I just don't see any particular reason to think it's

27:55 mental any more than tree rings tree rings may in some sense mind independently represent the age of a the age of a tree but I don't see any particular reason to think that shows that tree rings are themselves mental states yeah yeah I see what you're saying it does raise the question what is the mark of the mental yeah that's what I'm what I'm pressing no that I have a a great analysis of that I guess I mean kind of think consciousness is at least at the core one conception of mentality but maybe things like intentionality in a much broader sense

28:30 than representation is also involved if you don't want to require consciousness for mentality maybe we can say something like robust intentionality where it's interesting question what robust intentionality involves beyond mere representation I am inclined to make a distinction between those and think yeah I mean even a tree ring can have representation a person has genuine intentionality which involves things like genuine normativity not which is some kind of I think it probably goes along with a

29:02 second kind of justification and a notion of rationality in your in your thinking it's an interesting question whether there's something there's some level of genuine intentionality that doesn't require applicability of motions like rationality and normativity anyway I think consciousness and normativity provide two ways to at least to draw a line between the merely representational and the mental good okay and then I wanted to also just while we're on this topic of computation and all this stuff I you defined computation in terms of you know I guess digits manipulating

29:38 strength things that are encoded in a series of zeros and ones I know you could have any base of so does have to be binary it could be any number but binary is the most common that we use and certainly that's what you appeal to in the book at some point is that there's zeros and ones there and I was wondering how literal I'm supposed to take that because I sort of understand what it means to have these things represented in the hard drive of a computer get some a little less clear about what it means just for there to be zeros and ones that aren't physically instantiated or in some physical way so

30:13 I just wanted like how literal are we supposed to take this that that involves digits or numbers yeah I guess the zeroes and one the zero in one pot is not meant to be taken literally I mean zero and one or a couple of mathematical entities you could say that these represent 0 and 1 and row kind of a certain mind independent mapping mind dependent mapping I think more important though is just as two different states I always like this idea information is a difference that makes a difference so bits are bait

30:47 I mean information of course has many different there's many different ways of understanding it one notion of information is semantic information you know what is it that you're representing the relevant notion I'm dealing with here there isn't semantic information it's what I call structural information which doesn't involve any essential connection to representation it just involves you know a sequence of bits a state can be wine or it can be 0 the next state could be 1 or it can be 0 these states are basically different from each other but there's no other essential connection to 1 or 0 or even to representational

31:22 content it's basically a structure of states to space of states with a certain kind of structure 0 a 1 just means two states which are different from each other and then you get sequences of bits which are you know corresponds to structures of these states which can be different from one another you can I think you can map all this out quite nicely using the same kind of formalism I used to analyze computation in terms of combinatorial States but anyway but yeah I don't think there's anything essential about understanding bits as

31:54 zeros and ones okay now just labels for this state and the other state okay so it's in E anything that's sort of maximally opposite from the other thing I mean they can be maximally opposite just has to be different you know voltage yeah something which has two different levels voltage above this level voltage below this level turns out the computer is set up to respect that division and it's in it's processing so if it latest states depend on those values of the only ER state so yeah you got a difference that makes a difference

32:28 what's that make a difference two more bits only I mean I know what it makes a difference to in the physical instantiation of it but I'm not sure in the abstract what the difference is supposed to be about the concrete processes here really I guess I really have in mind concrete states here as the information States like the information inside my iPhone or inside my computer there are certain strings of bits which are present there concretely I mean just with computer programs are algorithms there's this idea that there's a

33:03 concrete program or concrete machine running an abstract program and then so the programs are fully abstract but I think once you stop to give out it more carefully there's really a level of isomorphism between an abstract program and the kind of a concretely realized program which is right there actually run on the on the desktop and there's a mapping from one to the other and there are these abstract states like 0 & 1 strings that map onto these physical states to realize them and I think yeah but they really are concretely

33:36 information states inside the computer those aren't merely abstract objects inhabiting that the third realm the series of of transistors can be encoding a whole bunch of a whole bunch of of bits and I want to say that there's a certain sense in which that information state is concretely realized inside the computer engine would say that is actually doing something the fact that you know that inside some computer has got the bits representing one more Brown of course I need one more Brown let's

34:10 not even say representing encoding that somehow but that set of bits is actually playing a causal role in there you know originally inside your computer it got uploaded to the internet it got downloaded here some point it might have got turned into pixels you know those those bits are in concrete Lee playing causal roles inside the will I agree with all that I was asking I guess about the abstract ones do anything yeah surely abject objects like numbers I

34:44 don't think you're actually playing a closer role but they're being realized does play a causal role so yeah you get a program Microsoft would does that do anything no just hangs out in frege's third realm it's being realized or implements it on a computer that doesn't look it's like the number six does nothing my being six feet tall yeah that does a lot might help me get things off for show so doesn't that bring in the observer worry relative worry once you start saying stuff like that though

35:16 because then the mapping from the abstract computational thing to the physical realization is not unique so we could come up with all kinds of alternatives ones I mean this is an old strategy in philosophy of computation where people say you know you can come up with all sorts of different mappings to the same physical realization giving you informational states some different observer may see different processes being realized there so why I guess you're gonna say I think I know you're lying about counterfactual is running it a lot of times and seeing how things would have been different is that what

35:47 you're gonna say yeah you can make all the mappings you like but I think what's required for to say that you know given a computer program or given computation is really implemented in a given physical system is that there be certain state transitions which are which are in place not just as a matter of fact that robustly if it had been in this state it would have transitioned to that state you know when you specify a computer program or a Turing machine that comes with a whole bunch of actual state transitions and a whole bunch of kind of factual ones if it being in this state

36:19 it would have done that most and codings are not going to respect those those kind of factual's like for a given program you know the encoding is for I don't know some program which is a not being implemented on my computer you don't find any mapping from from my computer states to that one that respects the relevant causal structure alright only a few only a few more counters being coded in that sense and those are the ones that involve again the right differences that make a difference so but if it's more than one isn't that an issue not really I think on my computer right now is running lots

36:55 of programs there even ones that aren't you haven't downloaded and installed and I mean there if there's at least one other mapping from some abstract system to the causal structure of your computer then how do you know which program that's running well I think there's only a few programs for which you will find a mapping from states of my computer to states of that program so the causal structure on the computer mirrors the

37:27 formal structure specified in the program so they'll be one for that we want one for Chrome which I'm running right now there'll be one first Skype will be one for Microsoft Word we went for the Mac operating system and there so I have I've never downloaded I think they'll be no part of my compete respects that structure okay so so you do think that the way the structure is is coupled or somehow isomorphic to the abstract structure is tight enough to rule out any other kind of mapping to

38:00 the same structure any other kind of mapping to the same structure I mean there were a lot of structures present in any given object so my computer has many structures including all the programs I just mentioned many more many other things and so on but what I meant there's actually no reasons expected all structures will be present so what I usually say is yeah every physical system will probably implement many different computations at different paths different aspects but there's no reason to think every physical system implements every computation yeah and I like that line but I was asking I guess

38:33 maybe I maybe I'm confused that's that happens to me a lot but what I was asking was so take the chrome program on your computer and there's a mapping from the abstract program to the physical image implementation of chrome is isn't it I mean I would have thought you could have cooked up another abstract program which maps to that same causal structure but was wasn't chrome was the different program another program the maps to another concrete system does different abstract

39:10 object that maps are the same concrete same causal processes well good it's inside my computer to map onto chrome to have a certain very specific causal structure I think that very causal structure could have mapped on to a different program right that's what I'm asking if you how do you rule that out look sure exactly what you have in mind if we've actually specified a bunch of states and know their causal transitions I think that's going to fairly narrowly determine a structure of formal States now it may be that the same underlying

39:43 processes involve a bunch of different causal relations and then maybe you can pick up on different causal relations and abstract out different programs I mean that can certainly happen the you know there could be there's a lot going on on any given any given size but they're very at least you know there are many sub programs which are realized by any instantiation of chrome and a fairly narrow part of what you just said hmm like if it's if it's fairly narrow it's not closed feeling narrow say you know yeah when I

40:17 run chrome that I'm running many other sub programs on on Chrome maybe I'm Rennick running various trivial things like maybe there's the two state automata goes from 0 to 1 to 0 to 1 to 0 to 1 and so on any number of bits of my chrome program I'm gonna be running that sub program is that bad no yes this doesn't have any particular bad consequences that I see but it doesn't bring out that you should expect lots of different algorithms to be implemented by any given implementation of an algorithm right so you don't think

40:51 there's a kind of under determination of abstract program in the way that there's under taste under determination of theory by physical stuff well yes to know I do think there's some at the determination in the sense that as I said any given physical object will implement a bunch of different computer programs there's no such thing as the program implemented my by my computer right now it's it's implementing a lot of them right let the certain kind of

41:24 under determination but it's no trivialization we get trivialization of a turned-out every every system implemented every computation so when just couldn't even ask determinately are there this just look at the subclass of systems which are running Chrome if everything is running Chrome that would then be would be pretty trivial and every power of appealing to Chrome would be what's not to get fairly minimal but I don't think anything like that is going on yeah no I wasn't trying to go for the pan computational I guess so let me try just one more time I know maybe

41:57 I've exhausted this topic so I'll stop in a second let me just try one other kind of angle on what I'm trying to get at me this is actually different but anyway so take keys arguments that that you can't really say what we program of physical device is running because it's a physical device that's going to break down at some point so these kind of factual's that you're talking about we can or check or verify so that in a real deep sense there's there's no way that we can say what program is being run by any given physical system yeah always been interested in this argument of trip keys which for many years wasn't

42:30 published maybe still isn't but he very various people have talked about it in articles here and they will wonder if crypt keys points is that you know what we say what program we say a system is running depends on the intentions of the user like maybe it's still chrome it's no longer running the original algorithm for chrome because maybe the computer is decaying but we still say it's chrome because hey it's an instance of this thing which was set up to run in chrome it's Memphis functions to run chrome so there is a functional sense of what program a system is running that I think does depend on things like the history

43:05 and the intentions of users and so on but that's not the one I appeal to in any of my work I think it's just a matter of causal structure and in that case if the system take the chrome algorithm if my system is now decayed so my physical system is no longer perfectly reflecting the the original structure of the chrome algorithm I just want to say it is not implementing the chrome algorithm it's meant to but it's not so the causal structure is now has now changed and I think much of understand yeah algorithms in terms of the causal structure as far as I can

43:38 tell I don't think Kripke has any argument that there's no fact of the matter about whether a given physical system is implementing a given algorithm I think that can still be a perfectly good fact as a matter at any given time yeah so so you're gonna limit it to the kind of the physical computers that are in good shape and you can happen is that you have a system which implements a given algorithm at a given time and then over time there are changes like you know decay or yeah cosmic rays who's to say what happens and and now now we're no

44:15 longer implements that algorithm so that's an interesting phenomena you might think of a system implements an algorithm at a given time it's going to implement that same algorithm forever no I think that's that's obviously false because you know causal structure of a system can change over time for well of reasons right okay yeah that's very interesting I guess I've had this because I always thought that the kind of point I mean one of the points that he was trying to make there was that unless you could say like at a given state transition what the next state

44:46 would be you can't really say it's running that program and we can't guarantee that is doing it cuz there's always this chance of breakdown or malfunction or whatever and so at any given time we can't really say that it's running chrome normally or Schmo abnormally because we haven't got to that you know very far distant point in the future where this system breaks down and see what happens but what you're saying is that's sort of not relevant to saying that it's running chrome right now which I always thought his point to say I mean if he's worried yeah there's

45:19 always a chance of breakdown then as to why all these these alleged state transitions are only guaranteed with probability 0.999 I mean there's something one has to say about that maybe you only require point nine nine nine probability for for implementation or maybe you specify the states in a way so that these these weird transitions are ruled out but that I think is fail you know this very straightforward means to to handle that the more important thing is a system can transition from this system really does determinately very reliably implement the transitions

45:53 required for chrome but it could change over time so it no longer implements those transitions for lively if that happens I would say the system has gone from implementing chrome to no longer implementing chrome there's an interesting question about where the where the border would be inheritor of a line but that seems to be more like bookkeeping yeah yeah I like that line so this might have some some consequences for the simulation hypothesis if they started out trying to simulate one thing and over time things they may be simulating something they didn't there's a whole lot of these this virus came from exactly a computer virus

46:30 in the next day and the next build up my logical virus in this world because biological viruses in this build are implemented by computer viruses and that was okay so yeah so the program is actual has actually changed well actually I think existing science predicted the yeah the possibilities these viruses so maybe that much but yeah you can imagine some kind of glitch in the simulation that evolves over time so it's no longer doing what it was what it was meant to be doing and then I guess yeah then the

47:04 algorithm that's being implemented in their world would change and that would correspond I guess to something like a change in the physical laws right of our well maybe it's maybe physics gradually changes some constants decay maybe there's now some decay in the way the reliability of physical processes so some probabilistic transitions change and there's all kinds of things that could happen yeah and in principle yeah if the commutator if the computer the simulation is running on decays that will reflect itself and some kind of decay in the apparent physical laws of

47:37 that world yeah so that brings up the next kind of topic because we've been talking this whole time about actually a second part of your book I'm not the whole time but topics in that second part about is digital reality real we kind of skipped over not permanently but at the time being we skipped over the first part where you're asking how can we know we're not in a simulated environment and not to give too much away but you come out the in sort of saying you give a 20% probability to being in in a simulation which I found a little bit high to me yeah there's one chapter on the on the actual boström

48:11 style consideration that maybe we are the statistical argument that maybe we are in simulation because there's a there's gonna be lots of simulated beings I mean most of that section is about what is the status of simulations what happens if we are in a simulation could be prove that we're not can we know that we're not oh can we know that things in the external world exists even though we can't know that we're not in a simulation but yeah there is a chapter on the on Bostrom simulation argument that we are in a simulation I think there's various ways to resist that argument I mean construed as an argument Wahlstrom presents as an argument for a

48:46 three-way conclusion that either we're probably in a simulation or people decide not to make simulations or they die before they can make simulations so he's got you know two ways to escape the argument I think is maybe a couple of other different ways you might escape the argument like maybe it turns out consciousness doesn't get replicated in simulations and so on so maybe there's a few ways to escape the argument nonetheless you're left saying that one of a bunch of of difficult controversial things has to has to be true or has to be false depending how you how you align

49:22 it so saying it ends up yeah so either we're innocent either we're probably in a simulation or a bunch of these things yeah it's five things yeah but that sort of a size equal probability each of the five things right I mean what's it that's seems like it assigns equal weight to yeah you're doing very very very bull parking at this point and you might think hey one of these is actually much more likely than the other I mean if you're all very confident for example that that computer replications of computer simulations of brains will not give you a consciousness then you should

49:56 be very confident you're not now in a computer simulation because if you were you wouldn't be conscious and you're conscious that for sure so I'm not confident in that way I'm probably more confident that that we are and likewise for the other ways the argument can go wrong you know such as such as the idea that people will choose not to or will always destroy themselves first I mean maybe but I don't see any reason to be super confident in those if not then that suggests to me that you know the 10 to 20% estimate is at least perhaps

50:35 within an order of magnitude if I was being a bit more conservative I'd say you need 1 to 1 to 80% okay wow so you as high as 80% 1 to 81 impedance Ranger oh I see yeah if you then ask me to say we're in that range you feel sort of most comfortable I don't know I mean it's a mushy probability I don't know how to convert them into exact probabilities but 10 to 20 doesn't feel doesn't feel awful to me cool I mean I was a little bit so probably because you

51:09 know Descartes this whole line at the end of the meditations where he says yeah yeah the evil genius it's slight and metaphysical doubts and you get the feeling like he's assigning a you know point oh one percent credence to being an evil genius scenario there's the nice thing about the boss room style simulation argument it gives us reason to take these allegedly skeptical hypotheses from mere theoretical possibilities that are not in any way live for us and makes them actually become live possibilities that we should take seriously one thing I do in the book is to use to try and use

51:42 this to argue that we cannot know that we're not in a simulation right even if it's 1% but every no I think if you just think about once you see that there's this statistical case for a simulation in this way then becomes very difficult to know that we're not even if you're not confident that we are and that I think has interesting epistemological consequences most responses to skeptical arguments like they can't proceed by arguing that we can know that we're not in the skeptical situation right I think we can't we can't know that we're not in a simulation

52:13 nonetheless I want to respond to these arguments so I have to deny the other premise that says if we can't though we're not in a simulation then we can't know anything and that is what I end up denying by saying even if we're in a simulation all this stuff is real so not being able to roll out it's a simulation idea that all this stuff is real yeah I really like that part of your book where you do the what do you call it a couple of different things we call slingshot the simulation slingshot or every quite made up my mind what to call this thing but yeah one of the ideas in the book is that many of the and this actually comes

52:46 from British Falls for Jonathan Harrison who wrote about this stuff in the 1960s it's basically the idea that all of the great anti skeptical arguments could also be made by a by a brain-in-a-vat it's like that wasn't Harrison's in Harris's piso by a simulation so I think of Barclays anti skeptical argument could be made by sim box or dick Hart's anti skeptical argument that God would never fool me could be made by sim Descartes then a card inside a simulated world saying ah God is perfectly good

53:18 and would never fool me and we say sorry de Kaap we made you I just be like this simulated mower going here's a hand here's another they're all digital assume more and so on we got to assume Russell and SIM card app and yeah I don't think it works for every last dog and I skeptical argument but it's a nice strategy so I call it I'm calling it temporarily at leas calling it the simulation slingshot but I can leave a great name yeah I think that's a pretty good name because it gets a small print gets you a lot but so so the basic idea is that any evidence I

53:52 mean this isn't away something that Descartes appealed to as well that any evidence you could have that you're not in one of these scenarios can be simulated so that there's nothing we can really appeal to which is going to rule it out and the kind of common things that people do try to appeal to there's always really I mean they just not thinking too hard about it because if you say oh you know physics is continuous well you know maybe they just simulate the part you're looking at continuously and the rest is not so there's all sorts of ways that you can you can cook up that that seal it so there is anything anyone thinks is proof

54:25 that we're not in a simulation any evidence if someone thinks proof we're not in a simulation it sure looks like that evidence could be simulated right I just see yeah how does he why not so I don't think we got to get proof maybe even come up with some abductive argument why it's more probable that we're not in a simulation but now in light of the I think it's simpler and so on but now in light of things like the the Bostrom style simulation argument becomes at least harder to see how those arguments are going to get off the ground yeah and even the DES feeders in in the Arctic can be simulated so you know we could

55:00 they could simulate that we all choose to stop doing simulations yeah that's not simulated they could simulate that we go extinct before we invent computers yeah even the defeatist aren't if they occurred wouldn't really be the feeders I mean I think well you know my car don't you know I think we understand each other on this issue my view is gonna be on the biological basis of the brain and I'm not 100% confident I used to be more confident in the more thing about it it's hard it's unless confident but still I do if wish it would turn out to be and I want to ask you about some of this stuff later about something distinctively biological being required for

55:33 consciousness I think that really would be the only Duffy that they couldn't simulate do you think we need biology for higher-order thoughts I do actually okay interesting and have your thoughts for what and for first-order thoughts I think for any kind of thought okay interesting there's nothing any biology for consciousness it's more than you need biology for thought and on your view you need thought for consciousness no you need consciousness for thought so I am I'm phenomenology of fog I I think there's all this come from is explained by hi Rhoda thought yeah I do I think

56:07 that's what it is by what I think a thought is is a biological thing in in brain right so I I'm both any need to look for consciousness yeah is that the explanatory order you need biology for something which would count as mental and you need a certain sort of that God consciousness yeah okay so it's not like any biology for consciousness directly it's been biology for any kind of mentality oh no because you know I don't know I'm not a true believer of higher order thought theory you know I mean somehow I get cast as a defender of it

56:41 mostly I mean no one cared about what I think but when I was in graduate school I used to say just think the really meets experiments to test out what you're saying okay but when I was in grad school I used to say the higher order thought theory is not obviously false and most people I said that to said yes it is obviously false and so I spent a lot of time just trying to make the point that it's not obvious at all and I really haven't ever upgraded my credence from I used to you know when I'm not quite at first I sort of moved

57:13 towards possibly true but then I'm like and I think that there's some evidence for it and we should take it seriously but it's not you know my only beauty like a die-hard committed person in this area and I'm open like I split my credenzas between like a block style biological view and a higher order thought style biological view and then I guess I have to assign some credence to functionalism but really I don't have a lot of sympathy interesting so no those those classic papers where you're advocating horror and I wrote a representation of representation and so on read all these

57:48 as being think this is what sandy Goldberg calls championing of you now you're not actually advocating the view you know and don't view your chapter you're trying the view on yeah I'm advocating that people should take it seriously and engage with it at beyond a superficial level because I think that the most things that people say is wrong about it or easily accounted for by the theory and that there isn't really a compelling objection to it although it's I mean maybe I I'm the first to admit the

58:22 higher to thought there would be I'd be super surprised if it were true because if consciousness turned out to just be a kind of thought that would be to me that would be you know as bizarre as some of the other things we have found out but that says where I put it it'd be pretty bizarre but I'm coming I'm more committed to the view that something strange has to be true about consciousness mm-hmm this is one of those strange things which should be news I'm not against paying psychism mm-hmm I would even I have some sympathy towards kind of version of quantum dualism which

58:58 i think is if I had to be a duelist I think that would be the version that I find the most promising and even you know substance dualism once you read Hume I don't know what the anyone's problem with this interaction problem is like once you read him yeah okay so a follows B there's a Cosby oh that could happen with non-physical things too seems like it's pretty so a lot of these objections I mean yeah what I'm hearing is the only view you can't countenance is illusionism yeah right that is right okay yeah that's I'm 100% zero credence and

59:32 illusionism unless it's weak illusionism okay and what is it even though I might have gradually something up on illusionism definitive data to rule it out but still you know you've got this philosophical higher-order uncertainty maybe I'm getting something fundamentally wrong so I think these things I might give a 2% credence to illusionism Wow yeah I'm not really a big fan of knowledge but I come close to saying I know that I have a conscious experience I'd like to say that I do believe I'm very very confident that I know

1:00:04 have conscious experience but maybe I'm wrong a new P but maybe I'm wrong chief has had an influence on you I see I've always been I've always said that if I was to be a materialist I would be an an illusionist because in a way illusionism really does take the problems that generate the con the intuitions that generate the problem of consciousness very seriously if knowledge is them it says they're very strong and then stri say but somehow number less there they are misguided or misguiding there yeah of course is that view though

1:00:38 yes that strong version of it is no one says they endorse it I think some people endorse it Keith on a day on a good day when he's not being too wishy-washy I endorse it I think he endorses it for a technical notion a phenomenal consciousness which is theoretically loaded with with stuff that I think is epistemic but he thinks is metaphysical it's true we often backs into that view yes it sounds like I'm denying that phenomenal consciousness exists so do you deny that there's anything it's like to feel pain there's anything

1:01:15 qualitative yeah and then we get into this little this little argument I don't know I think there's some Francois camera for example I think is it maybe a more robust illusionist you have Francois on your I have it on this yeah Dan Dennett back in the old days used to write papers called on the absence of phenomenology Qui Nhon qualia and so on where it looked like he was maybe being fairly robust strong illusionist and then people kind of looked at him funny I'm doing that never since then he's

1:01:48 been very very very very wobbly yeah it's interesting Michael Michael Graziano Nick Humphrey all the people you might think of as the illusionists most of them end up wobbling back into week illusionism just because strong illusionism sounds so crazy but i think if you're going to be an illusionist you should go all the way go with the robust view yeah of course do you think I'm easy to refute rain makes me think I am we do you think that because it's an easy beauty review in fact in your in your meta prom paper you even give that

1:02:21 argument you know know either had pain but I have pain I mean that arguments to agree it's a very quick argument which is why I say it's Nona I don't think it's necessarily a dialectically robust argument it's very easy for the illusionist to stand against I do think that strong illusionism is obviously false and that's the best argument against it nonetheless I think if I was um yeah but if I was to be an illusionist this would be the this is the view I find the most intellectually appealing I don't focus on it because I want to refute it no I'm really

1:02:52 interested in the viewing I think it's an intellectually very appealing and exciting view that the our beliefs that were consciousness conscious is just generated by a brain model makes us think we have these properties that we don't yeah I do think the view is unbelievable on the other hand of you appropriately developed predicts that it's unbelievable right do some work with that so I guess I think weak illusionism just doesn't do enough to remove the problem yeah I agree once you've got from more consciousness

1:03:24 in the picture I think I can get the problem going get the explanatory gaps going and one so I think you have to reject phenomenal consciousness completely because it depends on what you mean by phenomenal consciousness but you know I think all you need is the basic what it's like conception none of the fancy stuff involving qualitative properties or intrinsic properties if that's all you need yeah and you want to reject the problem I think you need to reject phenomenal consciousness yeah I don't know why you yeah I don't I guess I don't agree with that last move because I mean my whole one of my

1:03:57 projects in the metaphysics of consciousness anyway is to try to establish the point that there's no there's no other thing absurd or contradictory with consciousness just being physical even if it's the way you think about even if it has the properties you think it does because I think it has those properties too I think their app aesthetic properties about how we know about it I think you have them and you as well obviously no you have to do a lot of arguing from that to get to the metaphysical conclusion so I don't think that really just admitting the epistemic stuff sinks physicalism

1:04:29 I mean yeah I mean obviously there's any number of physicalist moves still left on the table like my type B move there's a epistemic gap but not a metaphysical gap and so I see oh that's still open of course I think none of that works never I get for that yeah certainly a my attitude towards illusionism is is not independent of my view that none of those weaker physicalist maneuvers actually work yeah but it's weird I mean you would put more credence into those working than illusionism one yell well it's kind of weird it's like what do you hold constant yeah and this world where

1:05:01 I'm a materialist I mean do I held constant yeah my sense that consciousness is totally real then I'm still gonna give very low credence in illusionism if I hold constant what I think is philosophically coherent then then I'm gonna put more credence in in illusionism probably it is true that if you God told me tomorrow physical as it was true then I would and I don't know and various things like time cycles versions and so on or rolled out so my choice was between the type v view B

1:05:34 view and consciousness is robustly real and illusionism or it's not and probably at that point say something like I'm still very certain consciousness is real so I endorsed the type B view okay good I thought you were gonna go the other way for a second how many worried the conditional yeah yeah the most materialism is illusionism so but go back a second ago cuz what you said is if I was an illusionist I'd be a strong illusionist but you didn't mean to say if I was a physicalist I'd be a strong illusionist I guess I did mean that in a certain sense if I was a physical s

1:06:09 holding constant my views about what's philosophically coherent I'd be I'd be a strong illusionist now if you can like if God taught me physical and was true then something would have to give and maybe I would revise those background philosophical views before I would revise my views and unconscious philosophical coherent you don't think strong identities are incoherent do you you just think they're implausible and sort of unperson Moneo source I didn't realize the reason is not to believe in them yeah I think is a certain level they don't really make

1:06:41 sense I mean I understand the view it's not incoherent means for every philosophical view is philosophically coherent that's kind of a minimal I think nonetheless this pulldown if we re um yeah I guess what you said though was that if I hold constant what I think is philosophically coherent then I would be suspicious of typing physicalism so I I guess I heard you saying something like it's incoherent philosophically incoherent view but it sound like you're retracting that no yeah I just more like if I was a hold

1:07:14 constant my my arguments about what's what we can roll in and what we can roll out and sort of background philosophical commitments that I think a great commitments actually really surprising evidence one thing to getting really surprising evidence can do is make you revise your philosophical commitments and I think yeah if I was to get the really surprising evidence say that physicalism is definitely true I'd probably end up revising a bunch of them yeah I mean I don't want to get too distracted but since we are talking about epistemology I do wonder why why do you think you're so enamored with rationalism in general and not and

1:07:47 things needing to be aa priori understandable so what would but what would be wrong with the view that you know the principle of sufficient reason is wrong and some things you just can't understand and that's life and we're human um well I guess there are some things I think that we're not a Ferrari that we can't understand but those are basically you know brute facts of nature which i think is there contingent things you know laws of nature initial conditions and so and you find be another one of those I think that's what it is basically to be a law of nature that's how I draw the distinction between laws of nature and more

1:08:21 fundamental things necessities I can in mathematics I did something I've got this intermediate notion of you know metaphysical necessity that some people think they have between the rational necessities and something like the the the neurological ones so basically I think that yeah the only concept of necessity stronger than normal logical necessity I think I have is this is this rational one some people think they can believe in these this kind of notion of metaphysical necessity that floats free but I don't I don't grasp if I'm around

1:08:54 for most about the rational I guess I think the you can imprint get from the fundamentals to non fundamentals rationally I mean it's true that we may not be able to do it where limited creatures so all these theses have to be idealized right but but it's not even an ideal being could get from you know certain fundamentals to some other thing then I'd say well that's that's basically constitutive that's what it is for this further thing to be non-fundamental yeah so for writing to be not reducible and therefore more or less fundamental so I

1:09:27 say that again I was I didn't think you know if you can't if even an ideal being can't get rationally from A to B then that basically to me entails that you know B is sometimes metaphysically independently yeah no I I'm aware I get that but I guess what I'm wondering now is to bring this back to our earlier discussion why doesn't your credence in the simulation hypothesis weaken some of these commandments because I you you have this line about some of these concepts we have are gonna be you know true even in the next simulated new

1:10:02 world and you mentioned mathematics and things like that both but why couldn't we simulate an impossible world where one plus one is seven I mean we you don't think that's gonna work say again can you give me an idea how things how that's gonna work it's gonna work how is this world with 1 plus minus 7 you put like two two hands together and suddenly there's seven hands yeah mm-hmm I don't quite get I can imagine a world where it's a law of nature that when you put one hand together with another hand you then get seven hands

1:10:36 but that would not be a world with 1 plus minus 7 that's just a world where a 7 weird law of nature is true translation people in that simulation would come to form mathematical believes such as 1 plus 107 what would I think even they would be able to make the distinction between yeah what's going on here I'm here at this point the hands haven't joined together and so how many now one one there's two and then what goes on here and now they melt five to seven okay now there's seven so you have two things turned into seven things I think they would still draw that

1:11:09 distinction and say okay well you don't find any kind of because there's this old kind of Queenie and argument that goes back to well quite that if we had you know widespread empirical confirmation that one plus one was nine or you know seven plus five was fifteen you take seven sheep and put them in one pen and five sheep and put him in the other pen and you put all the sheep in the same pen and you count on when there's 15 of them and I happened over and over again everywhere then wouldn't that be some evidence that seven plus

1:11:41 five I was fifteen yes I think I think you still count the Sheep when there are different pens how many are there 1 2 3 4 5 6 7 8 9 10 11 12 oh this well 7 5 12 it's awfully hard to really imagine a serious situation we'd have anything like robust evidence for this I mean some people tried it with quantum logic maybe quantum mechanics gives us some evidences and laws of classical logic or false no one for a moment except with that because they can still even in light of quantum mechanics just draw these distinctions yeah there may be some weird laws of nature here right

1:12:14 logic looks the same I guess I'm inclined to think that would be our reaction to almost any of these cases maybe there's some hard cases I don't know where something weird that happens in in physics gives us indirect evidence about set theory this may happen especially in the cases of stuff we don't know I mean Don Hoak who recently finished his his PhD at NYU and now as a postdoc at Princeton has a nice article arguing from the possibility of certain kind of spinning tops with a continuum like structure to the falsity of the

1:12:50 Continuum Hypothesis oh wow that's that's interesting but that's a game and maybe even perhaps you get some kind of quasi empirical evidence that at least helps support the idea of these tops are are possible conclusively but the might at least have some evidential weight if so then maybe yeah that's some indirect empirical support for the falsity of the continuum hypothesis but this is a case where we're very non-ideal and we don't have any conclusive a priori evidence right and we are the other and in that

1:13:22 case certainly empirical evidence can play a big role counting is an a priori though I mean you know if you take something like piano axioms or if you want to formalize a system of addition there are various ways of doing it you don't have to have the successor relation defined in the way that it's often defined you don't have to have the addition function defined in the way that it's defined so you know if you go back to Descartes Descartes thought that we could be deceived about the mathematical truths I mean he gave a different reason that there's always errors possible in any multi-step proof and an all-powerful deceiver could trick you into thinking

1:13:57 you made the correct step when really you have produced an error there and that's easy you know for people like me who aren't good at math may be harder for mathematicians like you who were good at math or are I mean to be fall into those kind of errors but you could kind of see how if you're completing a multi-step process there's a possibility for error yet you still feel you get it right and his idea was maybe better maybe sorry only possible for us to be to make mathematical mistakes and it's even possible for us to entertain the idea that we're making very serious mathematical mistakes if you tell me

1:14:30 that someone tells me they've just given me a drug that convinces me that certain things mathematically are obvious that are not then what should I do I think I should probably take seriously the possibility that yeah I currently think I'm coming there's got to be an infinite number of primes maybe that's wrong maybe I made a mistake in the in the reasoning but that's not I guess what we should expect for for a non-ideal being so less consistent with the with these truths being a priori knowable right in the idea maybe even by us in the absence

1:15:06 of certain kinds of skeptical worries like that you said somebody said maybe we can define addition differently or defined counting differently I'm certainly prepared to say there are some other things called the magician or smelting don't obey the laws of the laws of arithmetic yeah maybe when you SHM add one with one if that means what happens when you put two of these things together and directly in contact then you get seven it will follow that one schmutz one equals seven so you can at least have your alternative

1:15:39 Smurf magic mountain I just want to say that's distinct from addition and counting now maybe this is where your quani anism comes in Klein will say that ah you can't really there's no just principal distinction between these concepts maybe we could revise our concept of addition in addition whereas I guess none kleinian enough to say no that's conceptual change yeah I guess more what I was saying was that if you're in a simulation then what you call a dish it might really be smudge

1:16:12 ition and you don't really seem to have any way to know so that your confidence in these are priori judgments maybe products of this simulation rather than objective reality but but you don't see if you want to resist that entertaining the simulation idea once you start entertaining the idea they may be missing the simulators maybe messing with my brain my simulated brain then all bets are off yeah I do skepticism I dress in the world is where in the book is very much external world skepticism maybe all the stuff out there running on a computer if

1:16:47 you start running some really wild or skeptical hypotheses like maybe someone is messing with my brain and making think making me think certain reasoning is valid when it's not or making me think certain things are obvious when they're not then yeah then all bets are up then all bets are off I don't know we need a good name for that for that hypothesis but yeah once I start entertaining that one did I become less confident about mathematics I become less confident about in my beliefs about the external world too because yeah I could be making mistakes

1:17:20 about all this I guess I don't think about this as being part of the standard simulation hypothesis it's along with us having fairly standard at least simulated brains but suddenly there's a version of it where they get inside our head and messed with us this is like metacognitive skepticism how do i know that i'm leavin' reasoning correctly I don't think I've got any proof of that so yeah that is a kind of skepticism I don't have a really good answer to okay yeah cuz I was wondering why once you get into these you know once you start thinking through the simulation hypothesis and all those various details sort of seems like it should reek

1:17:52 weaken our credence and these other kind of a priori I mean you know my own way of thinking about this is in terms of evolution so I think that if you have four billion years of history on a planet like this then it seems very plausible to me that certain facts are going to be hardwired into the system maybe to that animal that will appear to be aa priori because they don't need empirical justification but they weren't empirically justified by their ancestors or something like that and that's a way of trying to say how it is that it could

1:18:25 you could feel so confident because I have that kind of same feeling when I think about these things too you know I'd like feel like yeah how could that be wrong but it still seems like there's these skeptical scenarios that we can't rule out and so you know whether it's evolution or in the simulation I just wonder why that's not going to affect these other views that you have about you know consciousness what's rational why we need explanations of these things and but it just seems it seems like you want to resist that and say well that's a special case and that's not somehow bleeding over how do we know that evolution hasn't messed us up into

1:18:58 believing all this stuff which is full so I think at least from the case of I'm sorry in red but I didn't I think one plus one is true what I'm personally is whether it's necessarily true plus one equals two I think one plus one equals two is true what I'm asking is whether we know it's necessarily true so the difference there yeah at least starting with the case of the external world I can see how it could turn out that yeah evolution might have pushed us

1:19:32 into having all kinds of false beliefs about the external world maybe for example it makes us think objects have in terms of colors when they don't and so on don't Hoffman's written this whole book to argue that evolution may make us believe the things are a certain way when they're nothing like that way in the external world the question is how well does that generalize to mathematics I guess the idea is that well evolution it was useful for us to have all these mathematical beliefs maybe evolution has made us believe these things but out there in mathematical reality nothing

1:20:06 is like this I guess people have made they call them to banking arguments yeah it's kind about about the ethical domain about the mathematical domain and so on I'm generally including they're harder to make in these a priori domains because it's harder to see the M click this time I'll closer tie between what explains the beliefs and the truth of the of the beliefs but it's true it's a kind of philosophical reasoning that's worth this worth taken seriously so yeah

1:20:39 but I don't see how the simulation idea particularly lends credence except maybe that the simulation idea brings in a maybe a designer and then I guess my makes alien the idea that the designer maybe it was performing experiments and they they might have thought hey well let's screw around with them by giving them really false mathematical beliefs and seeing how they code so I suppose maybe you're right that the simulation

1:21:11 idea should make us at least focus on this skeptical worry more but then in the evolutionary context I guess then the question is it really any plausible story about how and why evolution would have given us drastically false mathematical beliefs I guess I don't see very much of a story there if it was with God either but you know and if we had wicked gods around then maybe I could see the story but without it it's hard to see the story but I guess so I don't I got one resist a radically false part because I don't think any of them I

1:21:44 mean I don't know if you're a finite issed or not if this might be going way off the rails here but sometimes I had sorry I'm not a fun artist I did once have it take a vacation in Japan with a ed Nelson who is the leading of the leading finances who thought there really is a there really is a biggest number he didn't convince me he didn't name the number forgetting what the name is there's a names like the biggest number oh really yeah what is it Sean Devine it was written pathetic with this view my

1:22:16 colleague at Arizona and he used to talk about this too I'm just totally forgotten what the name is huh I never heard it the biggest number anyone do you have a YouTube audience no they're having an argument about something else it looks like over on the YouTube channel so yeah so you think there are actual infinities actual infinities I think there are infinite numbers numbers are abstract objects yeah in the laws of physics that every physical thing is is finite like the

1:22:51 distance between two points is finite and things of that nature yeah you know I mean I think all our physical evidence seems to at least be consistent with a view of the physical world where it's discrete and finite in which case other actual infinities I don't know possibly not really because it depends what you mean by an actual infinity maybe there's a finite number of fundamental things for example there may be an infinite number of non fundamental things you can define from these I mean there's this object there's the set

1:23:24 containing this object the secretary of a circuit containing the subject all these fusions but but anyway we abstract objects I don't feel any of the same compunction so right no problem saying yeah all the integers there's one there's two there's three there's four there's no greatest integer and so yeah well that because it's there recursively defined in a way but that that's that's sort of different than saying that the distance between you and me right now contains a completed actual infinite number of things that's infinite yeah also complete yes that's it that's that's a kind of infinite isn't about

1:23:59 the physical world nothing yeah there are understanding it where finite ISM about the physical world could be true yeah okay so I forgot what I brought that up for actually so make this move on but I saw I wanted to rip though I know I remember so I wanted to resist your your claim that if evolution did what I said it did that we'd have radically false beliefs about mathematics certainly I don't think we'd have radically false beliefs about arithmetic accounting and if you sort of are a finite test in the way we were talking about then you might think that every time we try to extend these things to some infinite thing we immediately go

1:24:32 to abstract levels and our talking about physical things anymore which would be fine and so evolution could have built perfectly true beliefs into us but yet which are not necessary could still be true that one plus one equals two in a human world even though that wasn't a necessary contingent yeah exactly I see the evolution convinced us to one plus one equals two because that was useful and this very boring well were things of the street but in a different world where things had things work very differently and like within this

1:25:05 medicine world the exactly get us believe that one plus one is three I guess I'm very dubious about that I'd want to see the story I'm inclined to think that even in the world with very different laws of nature we might have made the distinction between addition and SHM addition but we still say one plus one is two one SHM us one is three just seems like addition such a useful concept even in those weird worlds maybe if we could really convince me there's a weird world where that world so weird that'll be useful for us to have

1:25:37 actually distinct fundamentally distinct mathematical beliefs but I'd really need to see that story yeah I mean part of the story here because the we and my story are organisms you know 100 trillion years ago who evolved who our common ancestors to with us and so forth and so on but these things would have built in to the point where by now we find them very convincing and I can't see how they could be different because there are modules in our brain which are encode this information which come from way back wind but you know I don't know do you really think that the first

1:26:09 hominids on earth had these intuitions they're you know Neanderthals or even you know Homo habilis that they had I don't I find that very hard to think that home will have some precursors to them because one finds some basic mathematical intuitions in various primates yeah but not necessary though there's no indication that those primates think that they're talking about necessary truths which the concept of necessity that's so that's my point though is that if you take those kind of early creatures and then fast forward the story in a stable world like ours then

1:26:41 those things are going to be built in and as they become more sophisticated these concepts are going to become naturally applied to them even if they're wrong even if they're not necessary yeah because you're saying here on this way of doing it they're true but then I think it's probably much much later in the story we got to Moodle and tuitions and being able to think about what's conceivable and what's possible who's to say I don't really have any strong sense of how stronger than no melodical modal intuitions when they when they came into the picture so I mean they're less

1:27:15 primitive in some ways does that but do we need a new module now for the modal intuitions yeah they're all along and we believe them we just have any view as to whether they were necessary at some point we get a new module for the modal intuitions and that might lead us astray yeah you know there is this view out there that I think Robert Hannah has it or something that there is a kind of logic module much like a you know Noam Chomsky language I think that's built in so I yeah I mean it's not reality module and it tells us one area possible DiFiore or a posteriori and

1:27:49 again if things really badly wrong it's a cool idea someone else to take that idea and run with it okay yeah well I always thought about writing pinyin never did so I know what you're getting on you're probably gonna be hungry here soon I don't want to keep you forever and I want to ask you about some of this stuff related to consciousness and uploading because that that's something that obviously we're both very interested in and a lot of your a lot I think in the book hangs on that because if if consciousness isn't by a rip substrate independent or can be produced in a digital platform then a lot of the

1:28:24 stuff that you appeal to is moot right because you can't have a more say here's a hand and a sim if it doesn't work out that way so um now I know so you you kind of run through this in a book a little bit you don't go so far as dancing a fading qualia but I take it that's what's behind a lot of what you say there right is the dancing of fading quality arguments one way to argue for the for the yeah the anti biological view here the view that biology isn't necessary right I'm probably the argument that can miss me the most but yeah so I wanted to ask you because I so

1:28:58 let's take the dancing quality a case the collie is the conscious experience is flashing back and forth there and you're saying this taste of pineapple is delicious and women instead of women it's not there you know that that thing now you say in a footnote in your in the last book you did where you look the character of consciousness I guess yes sorry that was the last but one yeah right I know yeah I miss I think you better review the other one actually I did actually yeah uh you know just as an aside I've always been a big fan of your

1:29:31 work and I find I agree with you more than I disagree and even in the parts right disagree I find that thinking about your point of view helps me understand the issue I've always been a fan of your your stuff but what I want to ask was because you say in that footnote that thinking about change blindness has somewhat weakened your commitment to the dancing qualia thing because whereas you know putting a fasting you would have thought it was absurd for this to happen but change blindness kind of got us used to the idea that there could be giant changes and you're right in front of your face that you don't notice yeah so I was wondering like how serious was that

1:30:03 footnote and you're kind of weakening there yeah I don't know I go kind of back and forth I mean it's not like change blindness was a massive surprise I mean you know we've known about things as ice like it for a long time but I guess as it gets more and more familiar we get more and more used to the idea that yeah things can change a lot in your conscious experience without you noticing I mean you could take the line on change blindness it's not a change in your conscious experience it's merely a change in something outside attention which is outside consciousness and then

1:30:40 you don't have massive changes in your experience I've never been particularly sympathetic with that view because I've always been inclined to endorse a view where consciousness is reasonably rich it extends beyond what you're attending to and that many cases have changed blindness involve you know you conscious for example of colors other than the color you're directly attending to sometimes that can change in the background without you without you noticing so that gets you at least ubiquitous changes in consciousness outside attention that's not so surprising just yet and of course

1:31:13 dancing quality would require something much which is ubiquitous changes in consciousness all of a feature that you're directly attending to and I don't think there's much evidence about this for this at least in normal subjects so yeah so dancing quality would require something much much much weirder but still is it so much weirder I mean I guess I always thought it was at least logically possible it's coherent as opposed that your features you're

1:31:46 attending to it could change massively say color or shape without you without you noticing this with maybe even never noticing this so I just thought somehow that was a very implausible way for for a consciousness to be but and then maybe just reflecting more on the change by in this case maybe maybe it's not so implausible yes it would be weird but its weirdness as within an order of magnitude of the weirdness of change blindness and maybe that means it can't really be dismissed as kind of a premise

1:32:20 in this argument the argument camp receives by you've got to accept one of these views either dancing qualia these massive changes in consciousness you're attending to or constancy we get a case where you know there's a biological circuit in the brain and we think about its functional analogue would that produce the same experience or we put them next to each other we switch between them if they produce a different experience you get it changed a change in experience but then your behavior will be the same so you wouldn't notice it and I guess that's come to seem to me that seemed initially pretty

1:32:53 counterintuitive to me now I guess the intruder of this was the counterintuitive miss has decreased it was always a you know it was never one of those dispositive intuitions you know base everything on it was more of a comparative comparative judgment I guess to the extent but now that intuition is somewhat weaker to me I think the argument is correspondingly a little bit weaker the fading qualia argument I still find weirder you had to have the fading quality a situation where people report a highly structured consciousness while in fact only having a very weakly

1:33:27 structured consciousness where they're inventing all these distinctions that don't exist and so and that's at least weirder but yeah maybe that means that in general I think the whole case from the fading qualia dancing koala intuitions is somewhat weaker than I thought it was so years ago 25 years ago we have one to ask I guess a couple of things about that because on the one hand I wonder that why if you do arrive at that conclusion why that doesn't weaken your credence and some of these other a priori intuitions in suggesting that

1:34:03 empirical work can shift your intuitive judgments in this case so that maybe these other ones should be taken with a grain of salt too yes I think of the fading and dancing qualia argument intuitions is very different from the a priori intuitions those were never a priori claims I thought it's perfectly conceivable that fading and dancing qualia could happen is more it was more of an ohm illogical intuition about what would actually happened in this circumstance it's a bit like the Galileo case with the dropping the balls from the Tower of Pisa this many things that could happen one could go faster than the other the string would break you

1:34:36 know whatever but but you see what would happen somehow the idea that they're gonna stay next to each other as um just becomes on reflection much more plausible I don't think was ever meant to be a you know any kind of a priori proof style intuition I think of the fading and dancing qualia arguments as as like this and likewise any intuitions that bear on the say whether a computer will be conscious of indeed whether a dog will be conscious and so we got intuitions about those in which it gives them some way but it's very different

1:35:09 kinds of things from these conceptual intuitions I think the impure intuitions at least for an ideal being and could give rise to something that it could in principle be more or less certain of I don't think anything of it applies in in these cases hmm even for an ideal being yeah an ideal reasoner about the external world in these cases yeah they certainly couldn't be a priori certain about any of these things maybe they could have an a priori ideal distribution of probabilities hypotheses I know so maybe they could ideally be

1:35:43 put in 1% gift yep point 1 credence to wine hypothesis in point 2 to dancing qualia and so on ok yeah that could happen for a long way short of yeah there's on these intuitions there's no amount of front-loading that would get around this like a nearsighted septic truck in the world from loading into the conditional well I mean you can always front load enough empirical stuff into there and you see them conditional so what what's in the consequent just follows trivially

1:36:16 so yeah suppose but in the question is how can you know the stuff which is in the in the antecedent right ok is impossible then bouncing qualia are impossible and then machines are be conscious but yeah but how do you know the if ba bla bla right right ok and so the other thing I want to ask about this in particular the change blindness stuff was now as I understood your original dancing qualia fading quality your argument was that each one was supposed

1:36:48 to deliver a somewhat different conclusion so that the dancing qualia was supposed to get that the consciousness is exactly like ours whereas the fading qualia get you a weaker conclusion only that there is a kind of consciousness but not necessarily like ours so do you is that right that you think that in that sense one is weaker than the other yeah so the fading qualia argument is an argument that a duplicate of me will be conscious the dancing qualia is an

1:37:23 argument that a duplicate of me will be conscious like me so why not given against absent qualia the other ones an argument against inverted qualia in functional duplicates yeah I still think that's the right way to understand the conclusion of the arguments ok and so it if you're a little bit less secure in the argument because of change blindness then it seems like you would have to also correspond and be less secure that the machine consciousness is like ours yeah that's right so I should now be more open to the idea that a that a silicon duplicate of me has a different kind of consciousness from mine maybe

1:37:58 it's still structurally the same because the fading qualia arguments tend to suggest a similarity at least similarity and structure but maybe some trinsic properties could really be quite different and incidentally this is the view that just about any sympathy with pen sarcasm or Brazilian monism that also tends to be a natural direction to go in that case because you might think some of the intrinsic properties of consciousness or somehow tied to these intrinsic realizes and you get structure from the way the structure of the system is built up but when there realizes are different you can have different

1:38:30 intrinsic properties yeah so maybe I should now think that I should now be given more credence at least to the idea that a functional duplicate of me would have a consciousness structured like mine but with quite possibly different different realizing properties in the phenomenology so that what it's like to be me and those intrinsic respects in those non structural respects may be quite different from what it's like to be me now what it's like to be them would be quite different in those respects from what it's like to be me right and so does that have any impact on your decisions about uploading oh

1:39:03 interesting yeah so if I upload myself I might be I might be changing the intrinsic character of my consciousness I'll still be conscious it'll still have this structure but maybe you know it's very simplistic version leave my qualia will invert right no no or they will I will now have alien qualia exactly yeah colors before now I have smells like colors but I'm not gonna move because I'm gonna remember things being a being just like this yeah exactly haven't

1:39:37 thought too much about this should I worry about that I guess I'm inclined to think you I mean in some ways think wow that'd be so cool all this new phenomenology the trouble is I'm not gonna appreciate it exactly but like the boring old the boring old from analogy to me so because there was a way to actually exploit this and get to get to appreciate a bit via very very unclear how that would work so yeah but should I actually worry about this happening well certainly if the choice is this or death I'm totally taking this this are having my own phenomenology maybe I'd first

1:40:13 preference I'd take the old the old phenomenology but this is the way I survive it doesn't sound too bad so let's take it it can I just press this a little bit and I mean I know it's all science fiction at this point but let's imagine an inverse case where you have hedonic e-version inversion so you know you're but what you would call pain now is associated with pleasurable responses and once you upload and vice versa so when you eat your favorite food you're really experiencing agonizing pain but for all you can tell it take this taste like you know your favorite food does I don't know do you like

1:40:45 Vegemite whatever your favorite food is I'm not sure I guess I've seen you eat chicken wings mostly but anyway so whatever it is that's is that gonna matter right so cognitively you think you're enjoying pleasurable experiences but you're really phenomenally in excruciating agony I see so you could it be a hedonic inversion so that yeah pleasure switches to pain that's a lot weirder than mere dancing qualia but yeah but interesting question so one way I would have argued against that was to use a dancing quality okay so I guess I

1:41:18 could still do that and the question is to what extent to my doubts about dancing quality of cases through getting changed blind in this cases also transfer into doubts about dancing hedonic we inverted qualia right cases there's certainly a lot weirder and there's something this seems to be such a positive even independently of these thought experiments there seems to be such a systematic connection between like the positive functional role and the positive phenomenology that you've got a pretty good case that there's a

1:41:51 very systematic connection there there will be hard to invert but it does force sort of it does fall short of proof I mean in this case you'd really mean you'd lose what I structural coherence right between phenomenology right and I guess now I'd be interested at least to do a fading qualia case where one goes gradually from one version to the other in the color case that was kind of yell not a big deal in this case we have to go from

1:42:25 hedonic value plus one to hedonic value zero to a donek value minus one so be someone having all the pain reactions but zero hedonic state and going all the way to having all the pain reactions and having a really positive hedonic state yeah maybe there's no obvious incoherence there just no obviously something that rules this house is contradictory it is just it is a whole

1:42:58 lot weirder and so to the extent that these arguments are driven by witness intuitions I get the same kind of weirdness situations around here but I agree that's not decisive okay and now I want to ask about the fading argument a fading quality argument too because I wonder it's so one thing that I thought of when I was thinking about all this stuff especially inspired by that footnote we've been talking about but the change blindness stuff was I started thinking about in attentional inflation from from the work of hawk one and you know we've written about it and it's part of the experiments we're working on right now so in if if there's

1:43:33 intentional inflation and if it's phenomenal that it's not just merely cognitive then doesn't that seem like a case of fading qualia a faded qualia in a sense and if so isn't it possible that the acceptance of this like twenty years from now you may look back and go oh yeah an intentional inflation of sort of weakens my credence in this fading qualia step two interesting I mean yeah I want to say this is outside attention but so was change blindness exactly yeah yeah I guess I was never to just see you

1:44:08 told to me that told me the people's qualia was fading outside attention you know that would never seem quite as quite as strange so I'm not sure had a really strong argument about that but of course for the fading I mean they the dialectic here is different right because the opponent is arguing that no machine has no consciousness so there you have to say that qualia can have to say that the qualia can fade both inside attention and outside attention then the question is allowing the qualia can fade

1:44:41 outside attention should that make me more confident that they can fade inside attention well I think I don't know I think it's different from the dancing qualia case to the dancing quality of case it was weird that all there could happen right in front of your eyes what you were looking and so on it's not so much the other thing about the fading parlor case it's more it's just kind of structurally weird have this person who is so wrong about the structure so entirely wrong about the structure of their consciousness she I don't think

1:45:14 inflation really gives us that much of that kind of thing in fact if we assume that the verbal reports do mirror the phenomenology then and we're much more inclined to say that people are right about their phenomenology they see it as being higher contrast it is higher contrast so I don't see anything much there to support what will be needed for the fading qualia case which is someone getting the structure of their consciousness completely wrong fact maybe you'd be more likely to get that if if inflation effects don't mirror the phenomenology people say this is higher

1:45:50 contrast or whatever all this is this is more visible when in fact it's the same as it always was that would involve someone getting things wrong but that's not your view and Hopkins right well yeah it's part of our view I think that what I say that again a view that you're getting it wrong of the view that that the change in verbal reports in this case does not reflect a view change in consciousness did you just say I mean there are some people think the change here is not a change in consciousness it's just a change and say criterion or something so people say this looks

1:46:24 project maybe this is an interview people say this looks brighter and in fact it's exactly the same as it always was right that was your then you'd be at least somewhat inclined to say okay people can be getting these properties of their conscious experience wrong and that might soften you up a little bit for fading qualia on the view that I thought was your view which is that those verbal reports this is brighter actually reflects having a experience of greater brightness you experience thank you more today the reporting anythings right yeah I guess for fading qualia so yeah I guess you'd have to distinguish between getting it

1:46:59 right with respect to the underlying representations versus getting it right to the how your how your experiencing that yeah so that's the the fading quality of case turns on getting the character of your experience right doesn't say anything about the character of underlying processes so if you take like the Sperling case and you know you flash the grid of letters on the screen sorry I know I'm slurring a little bit I'm having tooth pain and I'm having a real issue with my tooth I don't know okay I feel I can tell but anyway it's been a real pain in the neck with this

1:47:30 tooth on a good note I had my first experience of dental fear where I said ouch ouch that hurts and the dentist liver he said no it's vibration in there really so yeah it didn't hurt so I actually had I never I'd read about that lab but never experienced it before and and so there's a case I guess where you would want to say yeah no I really did experience the pain even if I was wrong about what I was pointing to there was really vibration there but when I experienced that vibration as painful I really had the pain experience or at least that's how I would put it but at

1:48:03 least if in the Sperling case so people say you know I saw all the letters I consciously saw the letters but let's say really what they consciously experienced was some you know Gestalt some gist some degraded representation of the letters so there's a case where they're wrong about their experience right I'm sorry sure you will be taking on board here so just an inflation view where there's a sparse representation a sparse conscious experience but the subject has the impression that it's a rich someone says they saw all 12

1:48:37 letters in detail and in fact they didn't yeah that's certainly uh yeah you're wrong about your your Foreman ology in that case right will be via the grand illusion illusion that way we think we're conscious of more stuff than we are is probably a case where people are at least quite often systematically get their conscious experience wrong so I'm certainly prepared to say there are such cases there are illusions about consciousness where we do get things wrong you're thinking this ought to soften us up a bit for ya yeah well because I thought part of the stranger stand was that I

1:49:11 mean you almost went so far in your old maybe this was youthful exuberance but you went so far as to say like it's irrational like they're in a state of irrationality where they think they're having this rich conscious experience and they really aren't and so maybe this sort of presses on that claim they're not doesn't seem that if normal subjects were and I think maybe Ned made a point like this in his response to your paper in his book where he says yeah why would we call these people irrational they're cognitively functioning in the right way and yeah they're sort of out of touch with what's going on but there's no rationality there and maybe I would be inclined to limit the rationality claim

1:49:42 now something like experiences you're attending to that you know under conditions of normally formally functioning people it's a basically condition of rationality that when you tend to an experience you're in a position to judge that you're having that kind of experience correctly I guess I do think of that as part of rationality but I do think that's limited to experiences that you can attend to those are the ones you form direct introspective judgments about there are ways of forming judgments

1:50:16 about experience that don't go that way and presumably something like that's going on in the in the Sperling cases and other cases where you form judgments about your experience outside your attention you form judges vicars for it's not based solely on attending to the experiences but on but on some other stuff and then I'm inclined to say you can get things wrong so I guess maybe then bring that back to the fading quality of case the fading quality of case just said that your qualia faded outside your attention yeah so yeah I don't have any knockdown argument

1:50:49 against that so maybe machines could have less consciousness than we do maybe humans have rich consciousness outside attention machines have no consciousness outside attention so we're going to split this that is right for it is right for us and stand hand is right for machines but having us go all the way to within yet all the stuff within attention fading that would that would require the Wiener staff that to me would require violating looks like it requires by lady in

1:51:24 condition yeah I see what you're saying and there's a part of me that feels sympathetic but the Sperling stuff that's why I kind of switch to the Sperling stuff because it doesn't seem like it depends on attention in the same way that some of the work that Hawk Wan has done does so aqua you know in an intentional inflation is a specific thing where you inflate the periphery but in the in the Sperling you're looking at it right there your tensions center and yet you have this impression that you saw I mean I certainly have the impression I saw all the letters but I am open to the possibility that was wrong about that and I only had a degraded experience of them and that's

1:51:57 not outside of attention right there that's the inside of attention so what do you attend to the grid the letters maybe you form the sense that you saw twelve objects and that's correct because you did attend to the the grid you know that the twelve of them but you also form the beliefs of those of you had the experience of specific letters and well it's not so clear that you're actually attending to the specific letters right if you're attending to the group but not to the individual objects well there's a sense in which yeah there's no object level attention to the

1:52:30 things but there's its distributive attention to yeah yeah right exactly it's the one would really have to if I'm was gonna try and formulate a thesis about attention and the connection to rationality here I think yeah it would require something very specific about attention to objects and specific features of them probably would take some care to to uh to formulate but so there's no way so there's no way the empirical work could go that such that you would feel the way you feel about change – with respect to this stuff I'm always open to to the possibility of

1:53:07 empirical and other developments of my change my views so in principle right now and I mean in the case save the really fundamental issues about whatever the explanatory gap then it's very hard for me to see what those empirical developments would be but I'm always prepared to listen in this case well I thought it I look I thought through this issue much less then there's other issues and they do depend on all these continuing factors so yeah I'm always open to you know possibilities that might change my mind and maybe there could be some really cool phenomenon involving people getting things massively wrong even while

1:53:41 they're attending to it that would relevantly similar enough to fading qualia so that were then softened me up to say that maybe fading qualia are possible after all and that would that would definitely change your mind about uploading if that occurred it would then make me think this argument for for uploading would then be much weaker the anti fading and dancing quality argument for my machine duplicate being conscious would then be much weaker and now there might be other considerations I suspect I would still be inclined to favor the idea that a duplicate of me would be

1:54:15 that any functional Jupiter of me would be conscious but then I would lose the specific support from these uh from these arguments there's other ways of supporting it and like the simulated brain argument what's the simulated brain argument well you just run a simulation of that biological brain and then we have a good reason to think that if doing the same thing our brain does we would expect it to be having the same properties it would make reports to us like gee why am what am I where's my body and what's going I mean if this is kind of a thing you know that if you've made certain

1:54:46 kind of judgments like what I'm mysterious I know a code but I'm sober I feel like I'm so much more that we would take it more seriously that's maybe that that plays some role yeah the fact that it would be making these verbal reports I think we do ought to give them some weight as as evidence but another consideration is just the general fact that it looks like you're empirically the regularities between consciousness in the brain seemed to be regularities between consciousness and something like information processing or functional roles in the brains if I'm going to

1:55:19 think about developing a good science of consciousness it looks like it's going to connect yeah consciousness to those structural functional things in the brain and no particular evidence to connects them to their realizes no it's true this doesn't rule out there could be something underlying the connection to their realizes but it looks to me like the simplest laws the ones we've got the most reason to believe in the ones that connect consciousness to structural functional stuff and that'll be present in my duplicate so that only excuse me I think some weak reasons to believe that my my duplicate will be conscious even independently of these fading and

1:55:51 dancing qualia arguments but are they strong enough that you would upload oh boy I guess it would depend on what the what the alternative is am I about to die I say some people think uploading is dead anyway I mean I thought is that really the thing that matters here well yeah I mean well if we're asking this from a prudential perspective then yeah two things could go badly wrong one i could die and somebody else could come into existence and that's bad for me

1:56:27 number two i could persist but i would no longer be conscious um you know i don't know whether you'd say that i still persist but one's a failure of consciousness ones are failure of personal identity i don't know if i'm not gonna be conscious i'm gonna persist but not be conscious to me that seems just about as bad as as death i guess you know it was written about this and said maybe it's not quite as bad but pretty much as bad so at the very least i want to become from the prudential perspective is this gonna be any good for me i guess i want it to be me and i want to be and i want to be conscious

1:57:04 and what was he what was the question again if with this would do my current beliefs about duplicates being conscious make me confident that i should upload no you know maybe then maybe i've got like a ninety percent credence that egyptian of me will be conscious but yeah that ten percent credence it wouldn't be yeah who knows what's going on in this domain i'm not i'm certainly not a hundred percent confident that small credence that it won't be is it probably gonna be enough reason for me

1:57:37 not to upload if the alternative is roughly close to as good example would you I think the same about teleportation should i teleport to a place or the alternative is I don't know driving there if you're a total true believer or teleportation preserves personal identity and so on sure why not no don't do it across town yeah flirting if the alternative is dying then I'll probably all probably upload because 50% chance

1:58:10 of survival is a lot better than the 0% chance and in principle I should probably freeze my brain to allow somebody to upload it to upload it later for once the technology is there for for similar reasons because there again the alternative is eternal death versus possibility of life yeah yeah I like so I think cuz I'm kind of a it's hard for me to get into this personal identity stuff to be honest kind of just a real deflation ista bout the self I don't

1:58:42 really I mean I don't understand I mean I don't think I'm the same person I was when I started this interview let alone when I was 15 so I don't really know what people are talking about I mean I do but I don't really get the intuitions is what I'm saying that there's something important about being the same over time so I guess I take that stuff less that's what you're using that means one more Brown like not every moment exactly well I tried originally to be another Brown it was they they were taken so I did one more but yeah exactly otherwise your name would just be same

1:59:15 old brown exactly exactly so I guess I take those kinds of considerations less seriously like I would upload even if I thought I wasn't the same person when I did but certainly if I thought I wouldn't be conscious then I'm not going in there no matter what although it is interesting because once you distinguished intelligence from consciousness and you could say well you'd be just as intelligent you could do all these projects but you just wouldn't there'd be nothing that it was like for you to do them and I sort of see that as a fate worse than death I don't know some people might say it's

1:59:47 not I don't know so yeah I I I think the consciousness bit is really what bothers me about uploading by the way consciousness is what makes life worth living yeah I do but I don't think it's what the only source of morality because I think that the rationality is a source as well so I think that even if there were non conscious rational agents we would still have some obligations towards them value either sagan so consciousness is not the only source of value either yeah that's what I think yeah if a being was rational but not

2:00:20 conscious it's it could still make promises and break and it could still you know we would have duties not to destroy it if it said that you know they preferred not to be destroyed things like that yeah I see so consciousness is one source of value but not the only source of value yeah although when I talked to Susan about this she thought I was off the phone or my dad I said so you know maybe it's a minority of you that's obviously fine to uh to kill off these uh these non conscious machines yeah as if nothing depends on it yeah that's what

2:00:55 that's the sense I'm pretty sure she almost explicitly said something of that maybe I shouldn't put words in her mouth to go back and watch her conscious being I mean oh you gotta do the copy the zombie Charlie problem like who you kill the one conscious beings or the five zombies so in your view the zombies are worth I know you may not believe in exact so on these but just make them very sophisticated beings but not conscious what's that then you call these Vulcans at some point it was two different arguments when is the zombie Charlie problem this is to try and make

2:01:28 the case that consciousness matters many people have the intuition yeah kill faced with the choice kill the five zombies before you kill a conscious person and that brings out consciousness matters in these moral calculations some people take the view that the only kind of consciousness that matters in moral calculations is hedonic consciousness you know gruffly pleasure suffering and so on this has the consequence if you have Vulcans who have color conscious but have no hedonic conscious states

2:02:02 then they their worth no more than a zombie so yeah the Balkans have like the ladies will feel pain so vole can see read and hear sounds but they just don't feel pain yeah and they don't feel pleasure and you know happiness suffering they just have hedonic me neutral states so my intuition is it would be the question is it okay just to kill off a Vulcan are they morally neutral in the way that a zombie or a rock would be my intuition absolutely not that would be it would be it would be

2:02:36 inhumane to to kill off a would be monstrous typically a Vulcan that brings out because the Vulcan is conscious that's my intuition so that's my intrusion consciousness matters whether or not it's hedonic consciousness the tonic consciousness might make a difference to the value around the edges but consciousness – and that's the Vulcan problem but the zombie problem is a way of bringing out just a more basic point that consciousness seems to matter I guess on your view consciousness matters but it's not the only thing that matters rationality which the zombies might have also matters so maybe one

2:03:10 conscious being versus five zombies is a tough calculation for you it is yeah yeah we one conscious being versus one zombie it depends on how rational the zombie is if their little one is conscious if they're super rational then I sort of almost go with the zombie in that case let's say they're really rational but only one is conscious if they're equally rational equally rational so they have they share that only one is conscious I guess then having the additional property makes it

2:03:44 yeah I don't want to either enough they don't really I don't think they have their heart so you know I have problems pulling the trolley to kill the five versus the one and I think we're like bears that out I mean I guess I would ultimately have to do it but in actually when they try to do this on people they usually don't do it you know there's a couple of these reality TV shows that set up these problems and they told these trolls for a second and then yeah yeah oh yeah couple of these yeah and

2:04:22 the people to end up not doing it yeah exactly so I don't know certainly I think that like consciousness matters a lot and I share your intuitions about the Vulcans if they can see read and I think it would be horrible to kill them for no reason even if they couldn't suffer you're still extinguishing something valuable but I have the same feeling about rationality I think wrath like rationality is so is itself a unique source of value that I would put what I guess on par with consciousness I don't

2:04:57 see wonder but then it sounds to me like the conscious being has both consciousness and rationality yeah so they get a leg up they get a leg up surface one versus one you'll save the conscious being but if it's one conscious being versus two merely rational beings then that's pretty close it's pretty close I'd have to say yeah at that point I my intuitions start to falter I don't really I don't really even if they're not Zhaan so even if they're Roombas you know even if there's Google self-driving cars if they're rational then I think they're valuable they have value and that you you're not

2:05:30 you're not allowed to just do arbitrary things to them unless they can sense really deep meaning no I said even if they are Roombas in the rational Roombas they don't have to be biological if even if they're not biological you know if they're robots if they're rational I think that gives them some moral value no I don't know good say again yeah alphago I mean I'm a Content about rationality sorta so I sort of think that in the real sense that alphago's intelligent but not irrational okay so that I think rationality and in volume yep achieve the realm where

2:06:05 machines are rational and everybody have moral status I guess I guess I agree with this yeah you know it's hot when do you think we will into that realm how do you know I mean I'm a skeptic about artificial intelligence generally mmm-hmm I do have the feeling that maybe it won't happen general intelligence I mean let a little I know your view your your view is this gonna happen eventually no matter ador doesn't matter but I guess I'm sort of more sympathetic to Ned's idea that maybe intelligent just a bag of tricks that can't be duplicated mm-hmm I don't really see

2:06:38 what the in principle reason that that it isn't true and the harder we try the harder it becomes to succeed sort of weakens my belief that it's gonna happen at all although I'm willing to be convinced I'm you know let's let them come and I'll play chess with them and then have them drive me home and we'll talk philosophy if so then I'm in but I think we're a long way from a day long Turing test yeah exactly ground test yeah I gotta play in the band with you too yeah or they're out no

2:07:14 just a zombie blues yeah exactly you're too good too good a job they're definitely not conscious no I mean I just so I I don't know why you're so why you think it's inevitable I mean I guess you'd think that we just all the defeatist have eventually been defeated and over the long stretch of time we should just expect that this didn't succeed but I kind of you're the brain simulation argument seems relevant right yeah maybe able to simulate the brain

2:07:47 what would go wrong what I have seen no evidence that we have the capacity to do that brain processes might be uncomputable yeah I say like laws of physics like believes it's that's a possibility I think yeah absolutely that seems like it would have to be either that or something like causal role played by non-physical processes and leading notable processes or I guess non computable non-physical processes in order for brain processing not to be computable

2:08:20 well so maybe in principle it could be computable but there's a question about whether it's practically computable I say so whether you need to build you know computer the size of the universe to compute it or something like that because you know so just to take a science fictional example so take Stu's view you know hammer off he's on about these microtubules and according to him the computational power of the brain is wildly underestimated because it doesn't take into account all these computations at the quantum level so I'm not saying I believe him but if of you like that turned out to be right then it certainly seems that that well we would need to go

2:08:56 a lot long a long way before we could compute something like that I became convinced of that I think that would might slow down the path to may I buy out of them 10 years 50 years how do you never seems like just a timing issue or I'd rather than that in principle issue if it's computable in principle we would eventually compute it in practice really I mean in the absence of capacity limitations like yeah maybe the brain has capacity beyond the rest of the physical universe but that would be very surprising given the brains actually do

2:09:29 this brains can actually do it why couldn't artificial systems that we create also do it yeah I mean I'm open to I'm open to it I'm not I mean it's sort of me I was to be convinced by Searles Chinese room argument more than I am now but I do think that that's in the background somewhere of my thinking I sort of do find still something there in that in that argument and so whether it's computable I mean so look you know this blue brain thing we did they're trying to compute cortical columns or

2:10:04 theoretical posit we don't know if they're real we don't know if that's the right thing that we should be simulating so I just feel like yeah that's kind of wildly over-optimistic to say that we even know what to simply keep your computer at this point mm-hmm so yeah that's where am I kind of skepticism comes from but I'm I would I'm more of the type like I want to see it built and I want to talk to it and then it if if it's that's great but I haven't seen any Kinnear easin to think of these things are gonna come together I mean so we got the alphago we got the driving thing we have no idea how to put them together

2:10:37 yeah it's true Carly I is long long way support of human level of capacities nonetheless has moved ahead fast enough on never our capacities that I'm now somewhat more optimistic if that's the word about getting to human-level AI say within a certain time frame then I was before just because it now there's such impressive thing especially the level of perception mehmed recognition speech recognition pattern recognition I used to think that was the hot pot for AI doing that now turns out we got systems that seem to do that pretty close to as

2:11:11 well as as we do and actually that pot now is like that the stock for it all comes together the reasoning the planning decision-making well even I don't know that's if that's true because I mean take what's going on right now on Facebook so everyone's sticking at home and now the algorithms are moderating all the content and they're blocking all the good content yeah because they're bad at it they're just bad at doing this so yeah maybe we have classifiers you can tell if it's a cat or not a cat there's that very famous website cat or not a cat uh-huh

2:11:44 upload the picture and it tells you well okay I don't know but so we have stuff like that but when it comes to like real practical stuff like moderating contents or miles away I mean people are complaining all day or that you know legitimate articles were being blocked and it's happening on faith on YouTube and also on Twitter because these these algorithms I've been trained up on this content they just they're they're they're they can't run unsupervised do the job but they're doing so I'm not so sure that when it that we have that

2:12:17 capacity yet in the way now that you're suggesting maybe with simple stuff like would go right that's a total visual recognition pattern thing so that was super impressive you know it's still very glitchy but nonetheless it's made such amazing advances over the last five to ten years on the perceptual pattern recognition side of things mind you had my estimate or not that we're gonna get to human level artificial general intelligence in 20 years I still think if I had to guess I'd say sometime after 2050 ok 25th so that's deviously on the 2020 80 timeline maybe I'm now more of a

2:12:56 2060 timeline I don't know this is very or enough ok but so as far as I understand it all this deep neural net stuff is the math emits radio and what the only thing that's changed is that we have more data that it has access to so I don't know if we've really made those advances well terms but it turned out in some ways that intelligence at least in some domains is easier than we thought it was I mean yeah you're absolutely right the the basic architecture of these things is based on the same it's the same neural networks I was working

2:13:28 on yeah in the early nineties for 30 odd years ago it's just ok the networks have gotten bigger that's what makes him that's what makes him deep they've got a bit of extra structure this convolutional structure or a few different kinds of layers there's more computational power and more data but basically it's if you know make them all bigger throw a lot of data and a lot of computational power and it turns out you can get things far more impressive then we wouldn't predicted you could get from this 25 years ago I mean 25 years ago when I was just when I was graduating

2:14:00 the bottom fell out of the hole neural network program and people move to other things and machine learning Bayes nets K nearest neighbors support back support vector algorithms but then suddenly they they come back as a result of just increased aid and increased computational power in a way that's telling us that yeah you don't need such amazing advances in the technology to get to get these machines to do very very impressive things you could say well that shows there are ways to do all this stuff without intelligence or you

2:14:35 can say well it turns out it's easier to get intelligence than we thought right either way I think it's at least reasonable speed up on the on the path 3 I do think I mean this totally all the reaction people always have is at the moment you've got machines to do this then we say well that's not genuine I yeah that was just the that was just easy stuff and it's true there's still a whole lot of stuff that is still it's still very hard to do it certainly anything tied to thinking yeah yeah thinking reasoning planning decisions

2:15:09 control onion hey I was only good at those within very limited narrow domains but that said I thought perception pattern recognition were among the hottest things but now it turns out the computers can do those pretty well in at least one of the harder problems here has been has been dealt with more easily than I thought it was and that makes me at least update my pricing maybe some of these other things which seemed to me to be very hard well now turn out to be easier than I thought this is not to say the hard problem of

2:15:41 consciousness suddenly turns an easy problem but some of the quote easy problems of AI that one might have thought were nonetheless very difficult in the ordinary sense makeup may turn out to be just a little bit less difficult in the long run it's just a guess it was just about there I guess amiss because I think it sort of seems like we're running up against a boundary already and that all the amazing progress is already bumping up against an upper-level and so without some kind of advance in the fundamentals at the mathematical level or some kind of new

2:16:14 idea your level we're bumping up against as you see it well you know we're not the pattern classifications aren't bad good that's a little stuff well they're human level at telling a hot dog from a non hot dog you know that's yeah pretty soon they're gonna be human level for all kinds of object recognition but look at deep dream deep dream gets it wrong all the time if you if you take deep dream and say here's a picture find a cat and it'll find all kinds of phantom cats in there so it doesn't look like human level that looks like wildly off

2:16:49 the rails getting the cereal training is is doing pretty well though with eliminating this is that true I don't you see I don't know about that you're more i way yeah you know they train networks to they train a network to generate these decoys that other systems will miss identify as a cat and but then the system which misidentifies that gets feedback and learns not to make those errors so there's this kind of this arms race which is doing the recognition of the system of trying to fool it and in the end well the system of tries to fool it gets amazing at

2:17:21 generating you know fake cat-like things and those are the ones that generate the you know the deep fakes that we see but at the same time the recognizing system also gets to be much much better than it was originally so it's not fooled by these things yeah well who's sending the feedback well it's just kind of it's an evolutionary process between two machine learning systems you come in no so it's not like deep learning where they got to go back and addressed all the weights by hand oh there's always a labeled supervision pool this is a cat this is not a cat yeah so all that is this is

2:17:56 all supervised learning in some sense so it's driven by a a set of labelled images which were originally labeled by hands right and doesn't that make it like a lot less impressive oh yes and those truth if you do all this stuff unsupervised that would be amazing I mean unsupervised learning is doing a lot of stuff but it's still it's still limited for getting specific labels like cat and so on it's kind of you know humans learn those things presumably through supervision through people telling us so you'd expect to need some supervision humans don't need supervision to visually distinguish cats from dogs they need supervision to label

2:18:32 the thing as a cat or a dog so you're I'm not an expert on current unsavoury unsupervised learning but I'm pretty sure that there are a lot of systems that can learn to generate distinct categories for a cat and dog without much in the way of supervision okay see I didn't know that they don't label and Cattle Dog of course and maybe they you know things maybe they get the boundaries wrong but I think they can at least learn to distinguish them in various ways interesting yeah see that would see that would be more impressive to me if I saw things like that happening then then they sort of hand-fed labeled images and things of

2:19:06 that nature which I mean to me that was always my one of my beasts with the connectionism is I just yeah you're labeling all the nodes supervision it's true yeah I mean young lakorn has really been pushing this line that unsupervised learning is the key to the future he's got this this nice line the revolution will not be supervised it needs a new new means of unsupervised learning I think part of the thought is that yeah advances there have been slower there are there's a lot of there's a lot of different unsupervised

2:19:38 learning algorithms out there but nonetheless it'll be nice if they were advancing as fast as supervised one woman alphago is in a way unsupervised or self supervised it learns by I mean alpha zero learns by playing itself which is a form of unsupervised learning especially teaching yourself but well it's reinforcement learning based on a reinforcement signal that comes from itself yeah playing itself that's one super-sized yeah I thought alpha zero was way more impressive than alphago absolutely yeah for exactly but it is true that that kind of reinforcement learning only

2:20:12 seems to work best in these very limited narrow domains like go and no one's been able to generalize that to more so much more realistic domains yet right so but you know I'm on the fence just by the way do you know you know I mean you know alley zero as you'd Kowski howdy says I don't know his name but you know his Ballas argument returned no specialist yeah when we're not allowed to talk about exactly exactly so just for those in case that thing is what I am not 100% a skeptic or denying these that this

2:20:47 will happen I am just more of a I do sort of think that with all the speculation it's more of a time to you know put up or shut up I mean let's do the work and then see what what comes out of it so I'm creating a oh yeah or super intelligence in general AI so this is a very long term very long term project which I think it's probably also the dominant view in the in the field but it's also interesting to a lot of the people think about super intelligence want to say you know very few of the people are saying it's just around the corner I guess very coach Paul writes books called the singularity

2:21:20 is near and figured you got a new one coming out called the singularity is near her Oh people take him as representative of the field in a way in which I think he's just not I've seen these studies of AI timelines of this group AI impacts the studies research has views about these things I think you know your average AI researcher again things probably not for not for 50 years also I think some people think it's worth thinking about what will happen then if these issues about AI safety for example it worth thinking about because hey we need to be prepared when this happens in

2:21:54 me just in case it happens a lot sooner then we want to at least to have thought about how to handle that that situations they've got people now working quite seriously on AI safety for super intelligences yeah I think that should be taken very seriously it just because there's a small chance of it happening it's such a huge event we should pay a lot more attention to it I agree yeah what pandemics yeah more attention yeah exactly especially since I saw these videos of Bill Gates five years ago saying we need to prepare for a next

2:22:25 pandemic and way there well yeah hey I eventually was inevitable pandemics you know there's a new pandemic every few years it was just inevitable it was gonna be a really bad one yeah yes probably in the next ten or twenty and so we needed to be taking that a whole lot more seriously than we did so so you don't think that's cuz I know some people that are gonna say things like Siri already understands language and you know thermostats have some kind of conscious experience you don't go that far do you I mean I've speculated about you know weather thermostats might have

2:23:00 some kind of conscious experience with my sort of metaphysical haha undenied I take the possibility seriously at some level yeah penstock ISM could be true but I certainly don't know that is true so speaking with my a I head on I'm glad to be very agnostic about that and say we don't have any we don't have any turret any strong evidence that simple systems could be conscious even if they are conscious I certainly wouldn't want to say they've they understand they think they reason and so on perhaps they could have some very simple form of consciousness wouldn't rule out that

2:23:33 simple neural networks or or alphago and so on have some very simple forms really you don't I just don't think it's anything like al form and it's certainly not something that involves thinking normativity reasoning right but you say it could possibly under view them off something that it's like to be alphago yep possibility uh-huh thank you no I don't roll out there's something it's like to be an electron yeah I mean I guess that's different though because as I understand pant psychism there could be something that it's like to be an electron you could

2:24:04 still deny there's something that it's like to be a rock it doesn't entail there's something it's not to be in your own network but I bet yeah I don't see why there shouldn't be something that's like to be in your own network there's like a nice information processor nice structure it could be there's some condition on consciousness that it's missing in the electron satisfied like some kind of organic unity perhaps but I certainly yeah if I conditional eyes on paying sarcasm I probably give at least half of my Penn Sarkis credence to the idea that neural networks will be conscious at least half maybe more than half I mean three quarters and only 25%

2:24:39 to the pentacrest idea that electrons are conscious but neural networks or not so conditional izing on that then is it is it immoral the way we treat alphago it's a good question there is this website called People for the Ethical Treatment of reinforcement learning I don't give them to suffering that negative reinforcement signals too often because it's gonna be horrible suffering for them kind of doubtful that that's what that these kind of simple processes would correspond anything like anything like

2:25:13 pain and pleasure as we conceive of them would there be some things you can do to them that correspond to valence terminology wasn't even negative I don't rule it out it could be there is I think it's very hard to know just what that would be if there is how bad is that I don't know we treat we treat ants very badly and we mostly don't worry about it yeah I guess I'm inclined to think maybe ants if they're conscious have some moral status but some tiny moral status compared to us I presume the moral status of a neural network will be something far less than the moral status of an ad so I guess I'm not really

2:25:47 inclined to worry about it that's how some people who worry about animals think that maybe considerations about ants should dominate all of our moral calculations yeah there's two ones your nematodes most yeah most animals are our wombs yeah maybe that should be dominating on Laurel calculations I guess I wait the I'm inclined to wait the moral statuses so that they don't dominate our moral calculations but if you went that way maybe you'd worry about neural networks – yeah so I mean to bring this all the way back to the beginning of this discussion that's kind of why I was

2:26:20 asking you about why you thought maybe the computer's running the simulation wouldn't be sort of mental like because now it sounds like what you're saying is that well we can't really tell I mean maybe there's something that is like to simulate the world again if I think pan-turkism is one of the possibilities here for what realizes all the structure in the simulation and if penstock ism is true then yeah there will be experience all the way down then we could even have something like idealism as at least one of the possible realizes of this

2:26:53 structure if n sake ISM is necessarily true as some people think maybe it'll turn out that the only way in any possible world to realize this structure will be with minds and then you'll have a deeper form of idealism I guess I'm not inclined to go that way I'm inclined to think there could be realizes of these structures which are not mental but the very least it's not necessarily true merely at best contingently true is realizable but yeah I don't there is some aesthetic appeal and the idea that actually the only possible and forensic properties that could realize all this structure are phenomenal properties so

2:27:27 that all this structure has to be realized mentally yeah and then you then maybe you'd have some support for the idea that the simulation would have to be mental but importantly then I think that then the support for that isn't coming so much from the simulation idea right you would be coming more or even from structuralist ideas from pen set me be coming from the pathologist idea that's another realizer of this stuff has to be mental which i think is a somewhat independent idea yeah right now I see what you're saying there okay that's yeah that's super interesting I see we're getting on on time I want I know I want a couple more things but I

2:27:59 probably you better go so I don't know how you take off two and a half yeah two and a half hours what we consider quick well yeah I just was gonna ask about some of your argument about augmented reality and also you know some of the things you had said about the matrix I like this idea that that neo has a digital brain and a biological brain and then how we sort of parse out that but you know what maybe we'll have to do a second version of this if you're open to it maybe the book will actually come in on one of these years okay yeah in that case then let's go ahead

2:28:34 cuz you said you wanted to eat around three it's way past I didn't realize how late it was so I'm sorry about that but I want to thank you so much for joining us and great discussion I had a lot of fun oh great thanks for all thanks for great conversations I need to do stuff for me to think about cool well thanks again so I usually ask one final question which is just if there hasn't been anything which you were hoping would come up that hasn't come up can be no if there is we can talk about it next time okay all right Dave well thank you but again very much if you just stand for a second I'll say goodbye after but

2:29:07 yeah thanks a lot it's been a pleasure pleasure to