Consciousness Live! with Henry Shevlin (S3 E21)

0:00 yeah precarious at the beginning not yet uh okay we are live uh i was gonna say i feel like uh part of my brain is telling me because i'm we're interacting and we're not in the same uh quarantine bubble we should be wearing masks right now yeah i know right i realize of course we should you know it's fine there's there's thousands of miles between us but luckily they have not invented smell-o-vision yet henry [Laughter] we were promised that we were kids anyway so uh joining me today is a very special guest henry shevlin uh am i saying that right shevlin right javelin

0:31 perfect yeah i get a lot of my mail to henry sheldon um seriously like 50 it's like the standard there's something about the name that people want to swell swap the album the v but yeah you said it perfectly henry okay yeah i felt the urge to swap so i was wondering if i had swabs but so henry for for those who don't know you can you say a little bit about who you are and what you do sure so i think i had um maybe some slightly unusual pathway in philosophy i mean maybe not that unusual historical standards so i started out as a classicist so i did my undergrad

1:04 in classics 11 years of latin seven years of ancient greek and i did it at oxford which has a very permissive like famously permissive classics program and i ended up taking uh an increasing number of philosophy options and um whilst i was sort of you know they make you do a fair amount of a fair amount of the classics plato and aristotle the stuff that really uh caught my attention and captivated me was more philosophy of science uh metaphysics and then i guess the really big defining moment for me was um when i took supervisions with uh

1:37 tutorials with anita avramides in philosophy of mind my first real introduction to philosophy of mind um and i was going through a bit of a psychedelic phase at the time you know but i'll say i see wearing lots of tie-dye you know listening listening to a lot of pink floyd and um suddenly i got this i was introduced to the problem of consciousness and it honestly there's never really been anything any moment in my life where i've had such a rapid of also like a couple of weeks paradigms shift in the way i think about life the universe and everything i think it was basically a pretty much paid up um scientific materialist

2:11 um of the very kind of reductionist and narrow-minded type you know we've basically solved everything and then reading nagel and jackson and it really just made me realize holy crap there's this really important aspect of reality maybe the most important aspect of reality that we haven't really got a good scientific theory of yet we don't have a good grasp on and that one term uh has basically defined the path of the rest of my life um because this just completely captivated me um and so uh i then went on i did the b

2:44 fill in oxford where um i worked with uh tim bain on my master's thesis uh going into the b-fill i was really into sort of metaphysics of mind and was thinking i would uh you know work on the latest uh complex analytic analytic metaphysics of consciousness but um tim bain sort of i think very much in respect very sensibly sort of challenged me towards some of the more science of consciousness issues and in particular he introduced me to um a paper that was really that just been released um uh ned block's mesh paper consciousness accessibility in the mesh

3:17 between psychology and neuroscience and uh i was like oh wow there's this really interesting complex science of consciousness um that maybe can offer us some hooks some clues into this most profound difficult problem um so i ended up writing my master's thesis on yeah what is what sometimes called the overflow argument um so was it without 2007 exactly yeah so there were the two papers yeah so i wrote so there was the 2007 bbs paper and then the 2008 proceedings aristotelian society uh

3:50 paper both on sort of developing this uh overflow argument and um yeah so i was um really into that and uh wrote my thesis on it and at the time i was trying to find ways to make a denial of overflow more phenomenologically plausible um basically uh the developing idea i think in it's originally floated in one of the uh commentaries on blocks original on that original bbs paper by rick rush the study of generic nominology which has now obviously become quite a well established idea the idea that um you know maybe we're not conscious of conscious in the strict sense of many of

4:23 that many items at a time but they present themselves in these relatively generic ways so that you know i'm aware of a field of letters even if i'm not aware of a specific identity and this can the idea of this capture the kind of richness the appearance of richness of phenomenal consciousness right i mean so that's what i did for my master's thesis and then um i hadn't really anticipated moving to the us uh but um uh there was um i guess i became jesse prince visited uh oxford and um i met alva noe who would just join the

4:56 gc um the graduate center then and tim said hey you know look you've been in knox for quite a long time in terms of broadening your horizons maybe you should look at grad programs in the states and uh yeah and then um uh i ended up i did a quick visit to new york and got to sit down with ned block uh and say oh i wrote my thesis about one of your papers which was amazing and um yeah and and so i wound up at cuny which is obviously where we met um now and then at cuny ice which was i mean i had a fantastic time on the phd program at cuny um again working on similar stuff in the uh

5:30 cognitive science of consciousness philosophy of philosophy of going to science foster psychology i think cuny is a fantastic place for that i mean david rosenthal's cognitive science speaker series it's just like the bread and butter of my graduate life um yeah and it's really serious yeah it's fantastic and uh siri just thought i was trying to talk to her no siri now you oh siri [Laughter] series anyway sorry go ahead i don't know um i'm sure we're going to talk about ai as well so you know already it's making it self uh felt it's presence felt in the conversation

6:04 exactly so anyway i ended up um uh yeah uh doing a lot of uh of the similar kind of science of consciousness um type issues uh i started the philosophy of perception um reading group uh while i was there as a grad student uh pop shop which is uh still still continues pop rocks sorry pop rocks yeah still still continuing you're still uh doing fantastic um but i guess the other the next big shift in kind of my academic trajectory came sort of when i was sort of three or four years into the program um when peter when i started taking classes with peter guthrie smith

6:36 um and i'd always been interested in biology to some extent but around this time for kind of semi-unrelated reasons i became fascinated with mass extinctions and i still have this burgeoning interest in paleontology and suddenly i started to get the sense that you know maybe some core problems and philosophy of mine could be easily tackled by looking at um biology and i had some great conversations with peter and i ended up working with him on my dissertation and around that time he was also sort of uh beginning his project which is you know now there's multiple publications in a book uh come

7:10 out of it on um the evolution of consciousness yeah and um and again around this time i started getting super interested in animal consciousness um so my dissertation at the end was sort of a mixture of uh themes about animal consciousness and the science of human consciousness and um yeah and i guess the the next sort of big step uh was um when i came here to cambridge uh which is where i'm based now i'm sorry i should have mentioned that right emma i'm currently a postdoctoral researcher at the uh levy hume center for the future of intelligence which is a

7:41 fantastic ai um policy and theory and uh research institute um here at the university of cambridge and you guys have a podcast too right is that correct yeah when um i've interviewed dan dennett for the podcast which was which is fantastic um my um my former line manager marta helena uh left your hearing king but she interviewed peter godfrey smith um for that conference uh i'll put a link to it in the description after after this conversation awesome thank you yeah no it's um and so uh i should say a bit about the cfi center for the future intelligence

8:15 because it's uh i think it's a really wonderful institute and one of the few places like this in the world although so basically um i like any good philosopher mind um you know who'd grown up reading searle and foda and so on you know i had strong opinions about ai um but mostly kind of mostly kind of um uh founded in those classic 80s debates chinese room and so on um and although and i was interested in ai um but when this job um came up uh cfi and i applied for it you know i

8:47 thought you know maybe they want a special someone with a specialization specifically you know he's done five years of machine learning research so but um as it turned out one of the goals of the center was to get people who weren't already doing ai research who had different skills to bring to bear and channel them towards um doing ai research and the particular program that i'm affiliated with here in cambridge is called the kinds of intelligence program and basically the idea is to try and locate discussions about ai in a broader um a broader discussion about non-human intelligence you know

9:20 what it means and non-human consciousness non-human minds so that rather than ai being focused on a sort of this special little technical problem like little technical on this special technical problem it's part of a broader conversation about what makes human minds different how human minds different from animal minds and um you know and even with some practical practical sort of um real world implications um what way what are some of the ways in which animals still outperform even simple animals still outperform cutting edge ais um how did animal minds biological minds differ from uh from

9:54 artificial intelligence um which was so that which was perfect because i mean it meant i could bring to bear a lot of my prior interest in research in uh philosophy of congress science and animal minds to ai so i've spent the last three years um plowing into this exciting world and it is particularly at the moment really exciting uh wow yeah so that's kind of my life story that's a lot in there yeah um it's wow there's so many things i want to ask you now that that is going to throw every all the questions i had out before and and come up with new ones but

10:27 so i mean so it's funny you started out as a classicist because i was reading one of your papers where you cited plato as a as a the debates between the difference between perception and cognition so i was like i don't usually go back and cite the republic book 10. by the way book 10 is an extremely shady book of the republic most most self-respecting plato scholars don't read book 10 as far as i'm concerned um but so that must explain it right that's why you have a familiarity with those kinds of classical texts although you know it's funny my love of the republic came not in my uh seven years of ancient greek

11:00 but from a class with uh jakob vos vicilia at the graduates um yeah i mean i guess my attitude as an undergrad during classics was oh god these ancient philosophers look that greek is beautiful but they were wrong about everything i just want to read you know what bennett has to say and then it was actually later in life i started to go back i say later in life like i'm an old man yeah exactly a few years later that i sort of went back and said hey plato was uh dave rosenthal uh in particular the graduate center also pointed me said you know plato's write a lot about a lot of stuff maybe everything yeah and uh

11:35 these kind of uh positive influences i've went back and i i take a lot from um i think i draw i draw at least something from uh from the classics now from classical philosophers although i mean you know i wouldn't profess to be uh an expert in any of these guys really um but um there's enough of them in my sort of intellectual uh intellectual learning history to influence me i also sometimes think i don't know i'm interested in your opinion on this as well um you know this idea of the two cultures i think cpc i kind of think philosophy is an

12:06 interesting case because i think some philosophers are basically scientists and some philosophers are basically humanities people and it's not a simple it's not a matter of like oh the the uh the people who are scientists by nature do one branch of philosophy and then one two humanities people do another i think it's almost a difference of temperament and it's funny i think of myself although you know what i do is blush with cognitive science and you know i even even my dad my son has asked me daddy are you a scientist and i say well you know in a broad sense yeah i'm a cognitive scientist in some sense so even though i you know i what i do is science i think my

12:40 intellectual dna has like a strong streak of the humanities thinker in it um now what do i mean by that uh that's uh it's that's sort of an impressionistic description but i guess maybe one one thing that i found um is i think some philosophers and this is not this is not um drawing any value judgments here i think the more kinds of styles the better i think some philosophers may be coming from a more scientific approach they like to sort of take a problem they're annoyed by the existence of this thing we can't understand and they try and get rid of it in a nice clear systematic way

13:15 whereas you know um other people maybe this excuse more towards the people who have this what i'm calling the humanities influence they're like excited they're running around saying everything's on fire hey check out this fire here check out this problem here and i think you know there's a nice division of cognitive labor there um where maybe but if you have the more kind of uh problem solvers um not quite as simple as problem solvers and problem creators but you know those who maybe um focus on uh more kind of linear um mapping out theories trying to solve problems and

13:47 others who explore the space of problems with its place of solutions um but i mean i don't know that's just an impressionistic take i have but it's very interesting i do i thinking about my friends uh in philosophy most most of them would say he's more of kind of a science scientist philosopher he's more of a humanities philosopher but in that sense classics has definitely influenced me yeah no that's an interesting point i think you're right about that to a large extent um and of course i want to agree with you that we want as many kinds of philosophers as possible any anyone who's doing philosophy is great and i'm not big into policing boundaries or

14:21 i've never been a fan of like that's not philosophy and luckily i've never run into that very often um like certainly at laguardia where i teach no one has ever said why are you publishing in this journal it's not a philosophy journal or anything like that so that's that's good um but yeah you're right so i mean i have a degree in in science i have a master's degree in psychology and obviously degrees in philosophy so i i want to see both of them work together and there are many scientists who are basically philosophers they don't want to admit it but they basically are um and they do a lot of philosophic philosophical work under the guise of

14:53 you know theory building or something like that and then there are the philosophers who are basically scientists um but they're and i work with the people who are more humanitarian oriented um and who are more interested sometimes you see this split like in the continental analytic divide you know it's like continental people are more interested in like human problem problems that of where human fits into society and stuff but not always not always there are some analytic philosophers who are interested in the meaning of life and the value of consciousness and stuff so yeah i think that there is this kind of split um

15:24 but my my goal is to try to get them to talk to each other that's why i think that we benefit when the two types of personalities interact with each other rather than just with their own types and you know what i am and you probably probably maybe agree with me here i think i know one of my favorite uh uh accounts of what philosophy is is the seller's line about things things in the broader sense are things in the broader sense hang together in the broader sense of the term and making everything fit whether it's uh introspective reports uh kind of psychological notions about intelligence or value or consciousness

15:57 or desire making that psychology and making the psychology fit with the neuroscience and getting this kind of complete picture this holistic view of the world that's what i see in the broadest philosophy's job as being and for that you want people working to try and forge the connections of every every point in the web that makes sense right all the way from the really hard philosophy of physics all the way up to you know well not all the way up to areas like aesthetics or theory other aspects of theory of value yeah no i agree and i've always considered i mean i have i'm not a big fan of the seller's quotes although i

16:30 understand why people like it but you know i tend to think that philosophy has a kind of distinctive methodology that centers on argument um and so the counting argument is evidence as opposed to the empirical sciences where experiments count as evidence but uh um that's neither here nor there but however so i i do think it's and i tend to think of philosophers like good philosophers as kind of what i call modal knots they're exploring the possible space um the space of possibilities as it were and you know trying to i don't think we know much in philosophy but i certainly think we have a lot of

17:02 conditional knowledge like if this it turns out then you know pant psychism is ruled out or if this turns out this way then uh materialism turns out to be true so we have a lot of that kind of conditional knowledge i think absolutely i'm so glad you said that because i mean i completely agree um and i think another still on my biography one other aspect of my kind of educational development i think um well developments may be a bit of a normative thing like educational changes has been um i think as you can probably remember from my early days at the graduate school anyone who believes this theory is crazy

17:39 yeah i don't like that kind of attitude yeah i think it's surprising it's pretty common in philosophy i mean sometimes i think um it's a kind of uh instrumental dogmatism where the idea is you know you you've got to have people who will defend the theories the last but you know like a lawyer in court you know to make sure the client gets the best possible defense yeah you have someone who's completely committed to this theory and that will compensate um but i mean i don't know i i can't do that but i haven't i have a tendency what's the phrase right to break okay babe is that the phrase you know like you know be defending of you really

18:12 enthusiastic at some level i i'll say but of course you know i don't really know if this is true yeah i know it's ironic because just not to cut you off but you know there's that famous russell quote that people always throw around it's like uh the point of studying philosophy is to loosen the dogmatism or something that you know and but it's actually a lot of times it's not that way philosophers are very dogmatic and studying philosophy sort of just re-entrenches your already held antecedent beliefs you just learn how to argue for them like uh so the beliefs don't change very much but you see the arguments that can support your view and you kind of just stick your guns i'm not saying every philosopher is that way but

18:44 i i see in students at least beginners this kind of attitude emerge a lot actually it's funny and i don't think again i think this goes in every case i mean there are some brilliant people who spend their lives doing theory building and you know have very good reasons for holding their positions trenchant um but i do think very often the dogmatism is something you associate with the really good enthusiastic undergraduate in some cases i just find it um reminded of a very good friend of mine who's uh he was he went he's a world champion debater he went off and did compete in the world championships and uh you know i chat to him i'd say oh

19:17 um what do you think about the case for capital punishment or legalizing marijuana he'd say oh here are all the arguments for and here are all the arguments against i'd say okay what do you actually believe he said oh now that's a tough question yeah exactly um you know i ended debating with the university as well and i think the more you get familiar with the arguments on both sides in some ways the more the harder it is i find it to take clear views particularly a problem as big as consciousness um this was also something that came i think working through uh we're working with peterborough smith

19:51 but really has influenced me you know i think i um i in one of the early drafts of one of the chapters of my thesis you know i went through various theories of consciousness discussing why this theory is wrong that theory is wrong and he said look this is this is a too shallow this is a set them up knock them down approach you know setting up these simplified versions of theories there are people who spent their life writing really smart people who spent their life writing books uh multiple books on these theories developing them defending them doesn't mean these theories are correct but it means that you can't just casually dismiss them and in particular you know until you can

20:23 understand the motivations or why you think these people hold the views they do you certainly shouldn't be too quick to dismiss them and i think and that's great advice yeah that's that's great advice actually we need more of that i mean i always tell my students it's a lot easier to knock someone down than it is to try to figure out why a smart person would say something that i disagree with yeah and that really is that's really the work you want to put in is to figure out look this person is smart so why are they saying something i think is i think it's wrong but why are they what from their point of view is making them think that this is right that's a much harder project than just saying you know oh it's easy to say nitpick and you know i

20:56 was encouraging grad school to do that nitpicking too that's what half the seminars we were in trained us to do was just to tear down these sorts of things as opposed to figure out what was good about them or help improve the project yeah absolutely there's um an idea my dad did with me uh growing up something he called the click game where basically you're having an argument and that anyone if you want you could be arguing about whether to get indian food or chinese food for dinner like you click and then you have to swap sides and it was just a bit of a fun exercise um but it's astonishing how um how hard it

21:31 can be in the moment to sort of decouple uh use your position and recognize the good arguments on both sides and i think that that's a that's a skill that um yeah that kind of decoupling ability is something that comes that philosophy can can foster but it doesn't always when it's done right and i mean it's like for me i will get to consciousness in a second but you know i teach philosophy of religion and every time i do i sort of turn into an agnostic but as soon as i stop teaching it i revert back to atheism it's like my beliefs like start creeping back on me and then i go back and look at the arms oh yeah that was a

22:03 good argument oh actually that is a good point and then it like loosens me up but then it's amazing how quickly you just revert back to kind of a bass line and you forget the argument so to speak but you have to keep them in mind right this is something that you uh as sartre says there's no momentum in consciousness you have to keep reminding yourselves of these things uh because they don't just their lasting effects don't occur um anyway uh enough about you henry let's talk more about you i wanted to ask you about something that you said at the beginning of your introduction because you said you were a

22:35 staunch materialist but then you read nagel and so forth so so does that mean your views on the metaphysics of consciousness are in the duelist realm or like how would you describe your position currently um well it depends which end of the week it is i um i would say i'm broadly uh interested in scientific and naturalistic approaches to understanding consciousness um these days i've sort of think most of the metaphysics of consciousness is beyond my pay grades to use a glib phrase but i mean uh i i guess i am

23:08 impressed by the hard problem of consciousness i do think it is a really interesting hard problem um i think uh every theory of consciousness that i've seen you know it's like what churchill said about democracy you know she's the worst form of government that's ever been tried apart from all the others i guess i would probably say something similar about scientific naturalism and consciousness but scientific naturalism is terrible approach to consciousness but all the others are probably slightly worse [Laughter] definitely um i mean i think that particular shift was from a very kind of naive uh reductionist um materialism that

23:42 didn't even really have its premises made explicit i guess i didn't even realize that consciousness was a problem in our scientific understanding of the world and um i did have a dualist phase and also an idealist space but i guess the most my sort of setup position for most of the last decade has been that uh consciousness is a problem that science can make at least progress on and um has made has made some progress on and um our best best hope for understanding it is to continue that kind of project yeah okay i think that's a fair

24:14 assessment i mean yeah i kind of went through the opposite phase i started out uh when i first started school i was i was pretty staunch duelist although i didn't recognize it but then once i read descartes i was like yeah duh like who doesn't think this already um but then once i started studying neuroscience i was like oh wait a minute the brain's the only thing moving the body around so the causal problem really like converted me um to to kind of uh basically like an identity theory and i've stuck with that ever since although i do recognize that there are other that there's there's

24:45 some serious about pancakes and we're taken seriously i guess the only view i really have trouble with is uh um elimination or illusionism those sorts of views i really uh i struggle to understand and then once i do i'm like well they're not really a limitabilist or they're not really saying what they think they are so that's something yeah sometimes i feel like uh kind of you have this phrase modern daily arguments yes yeah so you know you have the outrageous and exciting and dramatic version of the theory uh which gets the headlines and then you have the kind of you know really defensible version of the theory that it's just kind of oh

25:19 okay fair enough and i think a lot a lot of the time illusionism sounds like that it's like quality don't exist wait what that's crazy oh well no i mean they don't exist in so far as these radical extreme views about qualia or false but when we talk about kwali we're really talking about this oh okay well that seems that seems reasonable but yeah i mean i don't i don't profess to say that's the totally fair critique of illusionism as a position and there are some really cool illusionist uh philosophers out there and i'm i think as i mentioned i'm a huge fan of dennis um [Music] one of the godfathers of that position

25:52 um but i do sometimes feel there's you know there's it's a less exciting or dramatic position when you look at when you get into details than it looks on the surface yeah no i i i tend to agree i mean i've talked with keith francis a few times actually and that's the sense i get from him and i'm going to talk to francois um by the end of the year he's sort of like a staunch illusionist too i'm not sure if you're aware of his work but uh so uh and but but dan dan dana um he's he's tricky i think he's uh not tricky he's cagey he's he's cagey he's very uh

26:24 i guess how would you put it opportunistic like he he takes certain ideas which he think can sort of get his point across and he uses them but i'm not really sure how committed to illusionism he is like when you really push it uh so anyway i i'm not sure you know i know he's a materialist and a functionalist of some sort but what's the illusion is adding to the position as except as a way to sort of explain why people have these mistaken beliefs um but uh you don't need all the fancy you don't need to deny consciousness exists in order to explain why some

26:56 people may have some strange views about consciousness anyway so i don't know i think he's i think he's a special case i guess what i'm saying yeah absolutely because i did notice that you wrote an article an encyclopedia article on qualia which i thought was a bit interesting actually yeah um so so because where i come from quality is kind of a bad word we use words like qualitative um you know there's qualitative aspects to consciousness maybe phenomenal being that qualitative states unlike quality aren't necessarily conscious

27:28 yeah exactly right exactly where qualia seems to carry with it at maybe you know not all of these things like ineffability and intrinsic maybe some people build that in but maybe that's not part of the concept i think we could debate that but it certainly definitely comes with this assumption that they can occur unconsciously i think that is uh a part of it so can i ask why you why did you write that article is it just because someone asked you to or did you have a vested interest and and uh yeah i guess um uh so this is this is so you said

28:00 this is the article i wrote yeah yeah so um so it's kind of an introductory article and i think um part of it is the goal um for me so i heard term qualia at first when i was going through that psychedelic phase i mentioned when i was taking uh um taking supervisions with anita from anita avermedis and oxford and i was like well i thought you were going to tell us you were taking something more interesting [Laughter] you might think that but i couldn't possibly comment uh but um and i found this word qualia i thought how is this not

28:34 part of how how do 10 year olds not know this word it seems like a fundamental aspect of reality you know we that we and we should all know this concept we all know this word and um and then you know as i said in more philosophy of uh philosophy of mind i realized that you know for very strange reasons i'm again curious into your thoughts on this phenomenal consciousness became the operative term rather than qualia um for the most part i mean obviously people still use the term qualia but it seems to be less people yeah p consciousness phenomenal consciousness phenomenality is my common place these days um but that word has a

29:07 certain kind of historical resonance for me um but yeah so when i was giving the opportunity to write a chat from qualia uh particularly to introduce people to this cool problem the one that motivated me so much uh yeah it's gonna seize that opportunity definitely so so generally though you you're gonna um you would say you think qualia exist that they're a part of conscious experience um yeah although i worry that uh i'm going to do the whole mountain bailey thing which is i mean look incompetent obviously qualia exists right like this is this dry cider that i'm drinking right now

29:42 there's something it's like there's a distinctive kind of flavor associated with that there's a um there are i mean and here maybe i'm gonna i can slip into the shared vocabulary there are qualitative properties that are bound in my experience that are very different from the kinds of more discrete discursive properties that i can identify in my thoughts or more reflective forms of consciousness but look when i say qualia maybe maybe i'd be happy to say that when i talk about when i talk about polio maybe i'm just talking about the conscious portion

30:14 of the non-conceptual representational states of my brain or uh qualitative states that i'm in um so i wouldn't want to put too much too much metaphysical baggage um on that um yeah i mean because one of the kind of pushbacks that i get from people who are anti-qualia um because i agree with you first of all by the way that there's a common sense notion of it where we're just latching on the the what it's like aspect of our conscious experience there's a sense of quality of which he has to be real or else we're zombies um and that i i

30:47 will i mean yeah death first before i say there were zombies um although i know people have said that anyway so the point though is that um some people who want to push back on this say look so why do we need this term uh what what theoretical work does it do that can't be done by talking about you know these other things that we've been talking about why do we want to introduce i mean part of your answer may be historical like this is what sort of turned me on to the problem or whatever but i am curious like from a theoretical point of view do you think it adds anything to have this term around or could we do consciousness science as

31:22 well without it and i think that's a really interesting question so let me fail to answer it by saying setting science aside um i think it's an important addition to our psychological vocabulary right i mean um if you read authors who are good at sort of trying to evoke the ineffable nature of kind of a bit of an oxymoron right but trying to capture these uh qualitative aspects of experience right the idea that there's there's an important distinction to be drawn between yeah these more reflected discursive verbal aspects of our experience in the one hand and these uh

31:54 yeah these sensory aspects but uh we don't have we find it much harder to put into words maybe are in some deeper sense in april who knows um i think that's a really interesting psychological distinction and one that um is implicit in you know the writings of people like bruce or anyone who's really good with uh sort of capturing the more sensory aspects of experience right um and so i think it is a useful word in the psychological sense now whether as we move from effects psychology to sort of scientific

32:26 psychology and cognitive science whether we want to keep that word with all its sort of uh introspective baggage i think that's a further question i mean i can see why um phenomenal consciousness and phenomenally conscious uh as terminology and radiology have become uh have seemed to not totally displayed but become more dominant than the phrase qualia in describing what it's like aspects of experience um but i think qualia are um a useful uh yeah as a useful concept for picking out this this one aspect of

32:59 experience i mean it's interesting when you look at the history of this uh because um you know austin clark who is a philosopher at university of connecticut and i spent a year up there uh with him anyway he wrote this really nice paper on the history of the word qualia just like how it's been used and it used to be used for for the stuff out there like the qualia where the redness of the apple and then you know now it is we use it to characterize the redness of our experience of the apple or something like that but it's just been i mean i guess one of the one of the kind of reports that you hear a lot from the from the people again on

33:31 the other side is like it's been used in every way imaginable so that it's like you there's no common ground you say it and people think oh you're talking about the same thing i am but really it's been used in so many ways that it's basically doing harm because it gives the illusion that there's a commonality that we're talking about the same thing which may not be there um and and i do sometimes feel the first of that worry i mean i'm not yeah so i do feel like that that's a problem yeah i mean i think there's a certain kind of um churn in philosophical vocabulary where

34:05 you get a new bit of theoretical vocabulary that comes along and people write a bunch of papers using it and it gets laden with increasingly conflicting associations and so the amount of ground clearing you have to do when you introduce the term and gets uh more and more onerous yeah and eventually people always get annoyed and say okay here's okay i'm gonna introduce a new term a new theoretical term and i just mean this bit and not that bit and then that's picked off and you get the same baggage being loaded onto that um which makes it hard when you you know you're trying to read i don't know yeah you're trying to read philosophy of mine

34:36 from the 1940s and 1950s through a modern lens you have to do this kind of translation work yeah the terms of the debate have shifted um and i'm not sure that's a great thing in the one sense but i completely agree with you but these terms get laden with more and more sort of conflicting assumptions about about uh about their scope about their remit and you've got to you've got to clear a path through that yeah and i like the term phenomenal conscious i know you know ned introduced that term in the 90s whatever but i think it goes back to kant you're talking about uh the phenomenon versus the pneumonia right phenomenal

35:09 consciousness just like the world as we experience it um is in my view an old way older way of talking about these things may be modernized by uh by the way ned introduced it but i don't think that he is like the originator of this idea that the world as we experience it is you know an important thing that we want to study so that's that's why i like phenomenal consciousness better because the trait has these kantian connotations that's nice yeah i i think i mean obviously another advantage well two other quick advantages one is that it it's more easily turned into well maybe not it's gonna say it's more easily turned into

35:41 an adjective well then again if you think of qual i mean one way of reading what the word qualitative is it's just the adjective we don't want to get into that but um consciousness you know you can talk about estates being phenomenally conscious but also it opens the door to things like uh non-qualitative forms of uh consciousness and cognitive phenomenology for example whereas i think i don't know cognitive qualia always sounded like a bit of a contradiction in terms to me huh but phenom about cognitive phenomenal consciousness doesn't sound like a contradiction that sounds fine i mean i'm not sure if you know it refers to

36:14 anything but um it uh you know anything other than the empty set but i mean it's interesting so really you're not a fan of cognitive phenomenology um i i i'm a supporter of it yeah of course i don't know i mean um i don't have a totally definitive view on this um i was and i realized sorry this is sort of a side note but i was really impressed by uh i really enjoyed peter brother's book um the scented mind yeah he precludes it there right now right i mean and i will say one um thing that that book really

36:48 brought out to me is just how uh how heavily laden all even seemingly dry cognitive processes are at least in my own introspective case with imagistic elements yeah um but i mean i don't know i mean clearly um we have conscious thoughts um but whether or not it's appropriate to talk about their how they present themselves to us in terms of the language of phenomenology that i'm yeah i i don't have a definitive position but i i'd consider myself like skeptically agnostic about um

37:21 congratulations i mean i find this debate interesting for a couple of reasons one is that we all have conscious thoughts and yet that there could be this kind of open question about whether there is such a thing like whether it's part of our our phenomenology or not seems strikingly bizarre because i mean it's like it's imagine that there are people going around saying yes i see colors but you know sound phenomenology what i'm not sure about that like we don't have that sort of debate there because if we take it that it's accessible um so if there is cognitive phenomenology then i think it's an interesting it's interesting to think

37:54 about why there could be disagreement like why what's what's going on i mean my own view is that it's it's so that the phenomenology attaches to the mental attitudes so that all beliefs kind of have the same similar um cognitive phenomenology and so we miss it for that reason because we're looking for the difference in the content right oh well my belief that two plus two equals four is doesn't seem different than my belief that paris is not in america and so they they're looking for a difference in the content there but i think no that it's different than wondering whether

38:28 paris is in uh america or whatnot so that's part of my diagnostics of the of this debate but i mean anyway you actually brought up sort of the mental attitude component there you know the wandering versus thinking because i think on the days where i feel more um well on the days where i feel more inclined and more open towards cognitive phenomenology uh it's very often if you're thinking about desires desires seem to me uh on one hand to straddle two worlds on the one hand like desires have a subjective clarity and immediacy they have they have something closer to this

39:00 thing to feel than you know dry reflective thoughts um but on the other hand desires are you know typically characterized as propositional attitudes um so i think they're they're an interesting case those that it seems to me there is pretty clearly um a distinctive kind of thing it is like um to at least in some cases to desire that you know to be craving a uh a nice glass of chardonnay yeah i mean people say i want it so bad it hurts right

39:34 i mean one of the things oh sorry go ahead i'll just say one of the things that really brought this home to me was thinking about atypical cases of beliefs and desires like in particular delusional cases um so cases where in the schizophrenic cases you know they may have all these strange beliefs but they don't really act on them i think the egan has a fantastic paper on this uh belief in belief imagination and self-deception perhaps i mean anyway talking about sort of the inferentially quarantined nature of a lot of these kind of delusions

40:05 right sorry is that is that the kind of exactly yeah yeah so my thought was that uh well they must not really feel like belief so they must feel different um than the way your beliefs ordinarily feel they they don't uh i mean my view is kind of flat-footed so i think that you know there's a sense of certainty or like this is truth this is truthful or this is like this is you know there's a sense of conviction is the way i put it in the paper i wrote on this but if uh when you believe something you really feel like convicted that is true um

40:36 and i yeah i guess i just find it you know plausible that cases of understanding like reading a sentence in a language you understand versus when you don't uh there's a difference there and so the understanding pit seems to add something to what's what's going on over and above the sensory component because essential components are the same i mean there's definitely a certain kind of feeling of insight aha moment you sometimes get um when you're solving the problem that i think i can that seems maybe closer to some of the kind of classic examples of qualitative stakes a certain kind of english of insight

41:10 um but i'm interested uh uh that you think sort of conviction or this feeling of conviction um and i feel like there's a bit of a weasel word there but this impression of conviction um is it guides what you believe because this is something i worry about a lot um uh what do you believe you know people ask you um people you know ask me what do i think about x and as in the example the debate my friend and my debating friend early on i find it really hard to know what most of my first order beliefs are and i haven't found a good way to do it you know here's a story uh you know when i was

41:42 seeing a therapist in new york you know she asked me so have you been feeling this week i was like that's what i'm here to find out right i mean i really don't feel like i have any kind of unmediated access to what i actually believe most of the time um and and even in the case of conviction i mean i i'm interested in what you have to say but uh it seems like i can get so um head up about trivial matters you know how terrible the last season of game of thrones was um

42:15 it's like it if you were just to go on sort of the feeling of righteous indignation i thought about that is that really the thing i care about most surely not and i would put many things way above that um and so i worry that maybe that kind of subjective sense of conviction is very up in a full guide sort of what people's most deeply held beliefs values are um but i mean i don't really feel like i have a dog in this fight other than a certain um uh concern about how we can really know what we really believe i mean i think very often we do get ourselves into a a place where what we think we believe

42:48 isn't really what we believe i mean what what what do you mean by that so like the games of thrones thing well i mean just give like a simple example i think we've all um uh probably simple example yeah uh someone um someone decides someone is madly in love with a guy who happens to be a vegan and this person convinces themselves that they fundamentally believe in veganism and adopted as a lifestyle but really they don't really believe in veganism they

43:20 just uh are doing it to impress this person but i mean if you ask them you know do you believe in veganism they'll sincerely say oh yeah i totally believe in it but i think it doesn't i think a more reflective kind of analysis of their personality of what really matters to them and what's what kind of attitudes are likely to be robust in the long run is going to say yeah they don't they don't they don't really committed vegans they're doing it there's for other reasons they don't really believe them yeah i mean sorry there are lots of different types of cases that's just one example the case where i think what we believe is what we really think we believe yeah i think that's a good i think

43:53 that's interesting i guess what i would say is uh maybe we're using the word belief in two slightly different ways since what i meant was just like a conscious thought yeah um and what you're seemingly talking about is like a like a deeply held um core personality trait or something that you're deeply committed to or something like that yeah and i agree we use the term in both ways and my use of the word conviction probably led you down that way of thinking but i don't think that's really

44:26 you know so i would say that person does believe that they're a vegan yeah because that's what's conscious that's what they're conscious that's you know what they're conscious of that's what they're experiencing um whether they have other implicit attitudes or unconscious attitudes which conflict with that uh i think is an interesting question but i mean um if you wanted to describe like what it was like to be them you would include like you know the the thought i i endorse veganism or whatever the case may be so i would want to distinguish between sort of the the shallow notion of just describing what

45:00 the conscious thoughts a person has is versus the deep notion of like saying what their fundamental like their core beliefs are or something like that yeah absolutely um so i mean i think uh the only kind of disagreement then would be sort of which of these whether both of these or only one of these really deserves to be cool to believe um yeah i mean i mean i i understand that it's a very frequently that this terminology of beliefs to describe conscious thoughts is very common and particularly if we're thinking in sort of more cognitive terms you know what police state are you eventuating at a given time but i think again looking at it

45:32 from a more psychological view it seems to me when we say what are his beliefs um uh do you believe in global warming you're not asking about what thoughts are going through someone's mind at the moment but more something like a more stable attitude yeah no i 100 agree so but i yeah um what i was talking about was just conscious thoughts uh not trying to analyze any deeper notion than that so but i agree with you yeah probably is it right to call those beliefs yeah probably not um except in this more like cognitive

46:04 psychology sense where we use that term but you're right i don't think the folk i think you're yeah so let me just say again yes i agree i mean i i still haven't found a satisfactory way uh figuring out what your core beliefs are um and you know all i can say is in my own case there's a certain kind of soul-searching it's a cliche i know but you know yeah um well maybe not maybe not in this election you know some other elections you have to do a period of soul searching which of these candidates best represents my values please i'm glad you said not in this election good yeah you know if you're

46:37 trying to decide between lots of really worthy virtuous people for your local council position you know now this guy's gonna prioritize education this guy's gonna prioritize public transport which of these really matters more to me and i still haven't i still i worry a lot about um i mean the role of uh you know uh motivated reasoning that creeps into our self-description of beliefs um i think that's that's uh yeah yeah no these are important points but i mean but you at least agree that you can tell like what thoughts you're having so you're not conflicted in that way in

47:09 the same way about these deeper like yeah you have i mean i think for the most part i would agree with that the i'll just take the kind of problem case that i do think about a lot um i'm a huge jane austen fan um uh i think she's one of the best literary psychologists probably the best literary i mean maybe shakespeare uh what about dostoyevsky yeah okay we've got the trifecta i um very often one of the one of the things that you see brilliant there's a great passage in emma and particularly

47:42 where she does this her characters are deluded about their own feelings systematically you know they think oh my god mr knightley is the worst but actually as an astute reading you can realize that this person is deeply in love with mr knightley and obviously one way to treat that is to say oh she's on she has unconscious desire for mr nightly but her conscious experience is that she she hates him she's annoying i'm not sure i mean i can see that's one way to sort of go about describing it but i think there's also a sense sometimes when we um go through periods of realization when

48:16 you say oh you know i was i thought i was mad at my friend about um being late but actually i'm mad because i feel he's not been spending enough time with me lately and you go through these periods of revelation you kind of get the sense that there was something there all along you were sort of dimly aware of right but you couldn't have vocalized the time and you would seemingly sincerely have seemingly sincerely have denied feeling had someone asked you know why are you mad at your friend i'm mad because it's late but actually at some level um went below the level of uh full-blown reflective awareness you there was there was a there was an

48:50 awareness of even at the time that there was something else going on yeah so those cases are interesting i think but but i guess i'd want to be a little bit careful here and say so we want to distinguish two cases so one um is the case where the awareness that you are mad at your friend for the reason that they're not spending time with you is yeah so you're consciously aware of that um and aware of it as that kind of state right like you're experiencing it as oh

49:22 i'm mad for this reason but it's off to in the periphery or it's buried other things but in another way you might say well you're aware of having that state but in the in the wrong way right um so you're aware that you have this state of anger about them not spending enough time with you but you're aware of it in such a way that uh it strikes you that you're just mad that they're late so it's the same state of anger but you're aware of it in a different way um and to me that second way sounds more plot like phenomenologically plausible it's like and then later you can kind of uncover it but

49:54 i guess yeah in those cases i find it hard to to from my own experience to sort of relate to the idea that yeah i knew all along what kind of state it was and then only later uh but it's so yeah i mean that may just be anecdotal to me so i'm not sure that counts very much well you know you're talking to a humanities person it's all anecdotal i'm just kidding um but no i don't know i think it's really important and uh i mean someone like jane austen has been a really big influence on me in terms of my understanding of work psychology

50:27 um so i mean just going back to those alternatives i mean i guess one question would be so to be clear that the view you want to endorse is that in some of these cases i was aware of it just under the wrong description yeah something like that i mean the obvious worry is right like i mean how wrong does that how long long can the description be before it's better to say that you weren't aware of it at all yeah well i think pretty wrong actually so it depends so here you're going to get really into technical the weeds about uh reference theories of reference um but i'm attracted to causal theories

51:00 of reference so that if the state is you know has the right causal link to it i'd say you're aware of it um if the concepts that figure in it uh have something to do with the state then i'd say you know so i don't know like you know to take the kind of classic kripky cases um how wrong do you have to be about you know newton before you're not thinking about newton anymore well i think he could be totally wrong about newton like not know anything about him and still be thinking about him if you you know have the right concept or something like that i mean so i think that same thing you could say the same thing here

51:31 i mean do you think that would allow me to say for example um confuse uh uh a belief for a perception but two very different kinds of states you know i think um i think i'm seeing something but uh i think i'm having a certain perceptual state but i'm actually just convinced that uh what i'm actually aware of is a belief and i just mistake it it's entirely the kind of whole class of mental state that it is yeah i think that's why i think that's possible i think that actually may be the right description of some some kind of disorders yeah i'm

52:02 yeah like i think um anton's blindness is maybe i was thinking of anton syndrome as one possible candidate yeah exactly i mean i mean i you know i think this is um i think we definitely do engage in these cases of misidentification a lot i mean another example is uh i think cases where we even confuse the mental attitude of certain states you know we believe this whereas actually we just want it to be true right exactly exactly i mean so these are these are interesting questions about folk psychology i wonder if i could ask you about uh so when it comes to ai for example

52:37 do you think so are these the kinds of things we need to have ai agents do as well because i know you endorse a kind of cognitive similarity view especially when it comes to like moral issues um so that we want to look for them the agents that can kind of do the same things we can do are these the sorts of things that they have to be able to do or it's a different set yeah i think that's a really interesting complex set of questions um let me just pick one aspect of that whole nice web hold on um so i mean one thing that i think is important is um this is something i've addressed in a paper on

53:08 um uh rich psychological terms uh these rich psychological terms and ai uh with care right yeah you exactly use rich psychological terms and ai with cab um which is uh basically this paper which i co-wrote with uh my former line manager uh helena um basically just point to what i guess you could critically say is a kind of sloppiness in the way that a lot of ai results are reported um by computer scientists i mean i think some of that is you know they'll talk about a system as being uh displaying creativity or curiosity or

53:42 understanding or goals yeah and look i mean when it comes to terms i mean there's a spectrum here right on one end you have terms like memory which we're very happy to describe you know uh computers as having memory right or even you know um recognition is an interesting term um uh we talk about um you know facial recognition right um but obviously recognition also has a psychological sense and so maybe there are some terms that are more where restricted specific uses in in computer science are more well established but i also worry that once we get on to terms that are maybe

54:15 thick in the normative sense right they have greater connections to notion like moral responsibility perhaps or moral patiency or under or epistemic states epistemic status things like understanding i um i think it's important uh for various reasons uh they're going to in the paper to use those terms of care um i mean i'll just mention a couple now um i mean one of them is i think i'm a huge fan of cultural science i think going to science is one of the really cool success stories right that's not moderate success stories of um

54:48 the 20th century uh science and it's really and particularly in terms of interdisciplinary projects i think it's fantastic um but i think there's a continual battling cognitive science to ensure that people aren't engaging in crosstalk and that people are using vocabulary in relatively irrelevant relevantly similar ways and so i mean that's one worry right that if you if it becomes totally diriger in um computer science do you talk about agents as understanding when in fact these agents are you know really way too simple then i think that's liable to weaken the project of cognitive science

55:20 and make it harder for different parts of the quantum science to identify using similarities training an algorithm versus childhood psychological development you know the more the more we can converge on a clean vocabulary for doing that the better um another example is i mean this is something i'm aware now that i'm there in an institute uh cfi which has a lot of interest in policy um public understanding of ais and even among quality makers is terrible i mean i think it's just astonishing i mean compute i mean you know i've been working this institute uh for three

55:53 years and i feel like i'm i mean i'm as i'm obviously a philosopher not someone who's had years of machine learning training i feel um very often humbled by the sheer complexity of some of the systems i'm trying to understand but i mean so plenty of people i've met who are making these decisions about you know what to legislate or against have zero understanding of how any of these systems work and so yeah i mean it's kind of understandable the way there's um i mean so computers are kind of magic um they present themselves as magical in a way that say an automobile doesn't right

56:27 i think there's a way of intuitively understanding roughly how combustion engine works there's fire and heat and pressure and it makes things move and the whole thing goes but um something like uh running um something like gbt2 or gp3 generating these how on earth does that work there's such an extra gap under intuitive kind of understanding that i think the vast majority of people don't possess um one of my favorite bits of uh i think it's in um i think it's in multiple publications but um he's got a great section where he goes through register machines

56:59 showing how you can build up through mechanical components to increase into doing logical operations and i've used that multiple times in sort of uh philosophy of mine classes to just try and demystify how computers work just just a little bit you know and even by the end of the chapter you're still a long way from from anything that could run halo stuff that modern computing computers can do no it's true there's so many working parts in there that it's the average person can conceptualize like you show them a turing machine for addition you can understand that actually but then how do you get from that to

57:32 what's what's the you know gpt3 is doing um but so so on your view then is it would be a mistake to say that like siri for example understands what i'm saying and that that would be an abuse of the term but it's not an imprintable objection you have right some future version of siri could understand in the real full sense absolutely yeah um and it's interesting i um again sort of connections here with what i really believe but i mean all the uh of all things that seem strikingly obvious to me uh it seems

58:05 frankly obvious to me that computers are going to be capable or at least in principle capable silicon computers are going to be capable of having the kinds of high-level psychological states that we attribute to uh to other people um i mean whether the actual computers that we think it's most useful to build will exhibit whether their behavior will be susceptible to useful explanation in terms of these same kind of generalizations who knows i mean maybe we'll find that you know the best way to or the safest way to build computers is um is to build systems that don't where these kind of way where it's just not

58:38 gonna be useful to describe them and saying oh it really believes x if it desires you know y's for the things being equal it's gonna it's gonna do something or other um so uh but yeah in principle i think it's it's entirely possible and i think realistically the sheer human desire for companionship to be seen by the other is going to lead us to build machines that do are deliberately um morphic in the full-blown psychological sense um but i think there's not going to be a big uncanny valley of systems between those sort of true psychological

59:10 agent uh ais and the kind of systems that we have that i think at the moment that really don't deserve to be described in psychological terms at all between now and then there's going to be a whole lot of stuff that's just going to be really weird yes like japanese care bots the seal bots you know something i think about in relation to uh yeah let's say or someone's persuaded really strongly by the chinese room argument or some kind of fairly radical bio-psychism so that they think you know there's no way a computer could be conscious or

59:44 have feel emotions or many relevantly uh mental states that we care about and i kind of want to say look even if that's correct history will bury you right yeah i think it's so obvious that well it seems so plausible to me that like 20 years from now 30 years from now people are going to be having full-blown deep meaningful emotional connections with these systems i mean maybe some people already do but maybe they're on the far end of like uh they're on the radical end but i mean i think it's going to become increasingly common for people to have these feelings towards towards ai

1:00:17 and i mean you know even if philosophers want to shake their fists at the clouds and say this is uh you shouldn't be feeling this i i think it is an inevitability pretty much yeah assuming we don't have that in the next 10-20 years well yeah that's yes exactly i'm going to talk we'll get to that in just a second but but i think you're i mean there are already people who like there is a documentary about a woman who married the brooklyn bridge um she's had deeply not just you know deep romantic feelings for the roman sexual feeling she says in the documentary for the for the brooklyn bridge so i mean people are

1:00:51 do this um the question is whether they are justified uh you know so you know people get attached to syria they're bomb bots in the military um you know there was a case of touch in case of some some soldiers who their ai got shot and they had a little funeral for it or whatever so i think this is just like this is something that we do but but the question is whether the i mean obviously i'm i'm somewhat sympathetic to biopsychism um but not in the face of fully like functionally equivalent you know like if if there was an ai walk

1:01:24 around talking like you i would be like i would be done you know i'd be even if you cracked up in my my brain and saw that i operated to radically different principles yeah if i could walk the talk perfectly you'd be sympathetic to that system's being conscious oh yeah 100 um i just i'm skeptical that we're going to get there uh and every time that we think we are we had a we hit a wall i mean you know there's that famous quote 10 years of working ads enough to convince you to believe in god you know that when everyone talks about it um so it's you know there's this you know the gary

1:01:55 marcus kind of line which is that maybe human intelligence a bag of tricks um and there's no generalizable principle that we can scale up uh yeah i i worry about i mean yeah you're right the it's it's an empirical question and maybe 50 years it'll be moot but i i do think at this point it's not clear that i'm not as optimistic about the path from here to theirs as you seem to be yes i mean the kind of debate between the likes of um gary marx and lacoon has been uh really interesting to watch um yeah and i think um you know uh

1:02:27 both both sort of sides in this sort of i guess civil war within sort of the ai commentary field i guess just in case any any anyone any lessons i'm so familiar with kind of what i take to be the kind of core dividing line here is basically are the kinds of systems that we're building now the kinds of systems that can be scaled up to full-blown human intelligence have we basically figured out what intelligence is in some broad sense and we can just bring bigger and faster and better systems and they'll do it or are we still missing some undemounted traits um i think that's kind of the core i mean it's a bit simplistic but that's

1:02:59 kind of what's at the core of a lot of um a lot of these debates and i would have actually for a long time said um that we're still missing something important um i think there's a fantastic paper um i think i'm going to get the title mostly right here how to build machines um uh that learn and think like people from josh tenenbaum's lab first author is uh as lake um just just go through some of the interesting ways in which current systems i think it's a 2018 paper

1:03:32 it's a fast moving world but it points to some of the ways in which uh current architectures seem to lack the potential to perhaps do the kinds of things we'd want to see in human level intelligence um but there's kind of another part of me um particularly seeing uh some of the successes associated with uh gpt2 and gpp3 so i'm just about the gpt-2 was able to learn to play chess just using and it wasn't programmed to do that or like wasn't trained on that kind of data at all well i mean all it was trained on is chess notation right i mean nothing no there was enough in its

1:04:05 corpus that you know i could pick up chess notation and then it could on the basis of just recognizing the kind of the kind of moves that most probably follow other moves was able to play a decent game of chess and so i mean i think one of the really interesting questions um uh fundamental questions pointing to science going ahead is uh is intelligence one thing uh i should clarify here sort of a more kind of human uh psychometric sense of intelligence i'm not so interested in that i'm interested in more kind of big picture intelligence in the sense of uh flexible problem solving something like

1:04:36 that is that is is everything ultimately something like pattern recognition prediction um or are there other uh fundamental techniques or other impr in practice necessary shortcuts uh or heuristics that you need that you can't get to human level intelligence just through being really good at recognition connection well that's a really interesting question um but uh i think we have a good chance of getting at least good data on over the next 10 years as we see these systems just get bigger and bigger and more powerful i

1:05:09 mean i was blown away by the transition from gbt2 to gpt3 yeah i mean one of the things though about i mean for people that know gpt3 it's just amazing i mean i only got to play with it because i did ai dungeon i don't have you played i and i mean it's just amazing um but uh you should go check it out immediately if you especially like me grew up playing zork i'm an old i'm old now so you know zork was like my first video game basically you've [Laughter] yeah i just re-found the version of it

1:05:43 online i've been playing it recently um before ai but anyway so it does amazing things um but the number of parameters like how much data you have to feed in how many parameters you have to twiddle with it's not to me indicating that they're on the fast track to something like human level because if you scale up like if you just sort of think about okay so how how many of these things would you what does it have like a billion or trillion or has some absurd number so the scale and it can do this limited thing but to get it to do things like we

1:06:14 do how many how much data would you need how many parameters would you need just seems unreasonable that it's scalable um yeah if you do some um back of the envelope calculations i'm not saying this is the right way to go on this but uh i've seen some people do these comparisons when they look at the total number of synapses in the human brain and compare that to the number of parameters in gbt3 and it's still a big gap i'm not going to quit the figures because get them wrong now but there is a there is a gbt3 is still a lot simpler judged by this crude metric than the human brain it has less parameters than there are synapses in the human brain exactly

1:06:46 substantially i mean people throw around this number but about synapses but really what they mean is possible synapses so i'm not sure which one uh which one they were talking about like actual synapses or like ones that they're capable of making something like that um so so maybe so maybe that's i mean that's a fair point so yeah so i think this is yeah um an empirical question um you know i mean i i don't think gbt3 is an architecture it's possible it's a language it's not got the moving parts you need for intelligence um what i did take from the transition

1:07:18 particularly from gpt to gpt3 was not oh look we're closer this is a key step in developing human level intelligence it was rather that maybe a lot of problems that looked like they needed fancier solutions could be solved through quite simple means and maybe that tells us something about the nature of cognition and intelligence such that um a system that did nonetheless work quite differently from gpt to cuban three um might nonetheless through one one or two simple tricks one or two simple techniques be able to scale up human level intelligence not that the

1:07:50 precise architecture of gpt's qpts is is going to give us sensory agents but maybe intelligence is maybe in there are useful ways to think about intelligence according to which it really is one problem is so is on your view is intelligence itself one of these terms we should use with care yeah i mean i almost think uh intelligence should be for the ai side of things i almost think it should be ditched um uh one one phrase that i really like

1:08:23 that i've seen increasingly people using is artificial cognition yeah um i think that can be a more helpful way of thinking about it and you can talk about the performance of artificial cognitive performance if you want to turn it into a adjectival construction um yeah i mean i think intelligence is town with so much baggage i think there are all sorts of quirks about how we how we use um and how we attribute intelligence it's heavily value-laden um yes so you know i mean so it's it's kind of striking it's uh something that's a friend of mine who's a teacher as

1:08:54 comments done recently and i think it's also been um discussed by uh freddie devoer in a in a new book um but how we put this massive value on intelligence such that it criticizing someone's saying saying a student is lazy or is under-motivated is totally fine from an educational perspective but you shouldn't no teacher have a call or should ever call a student you know oh he's just not that smart right yeah so i think intelligence has this really powerful normative um status

1:09:26 in uh in the way we think about education the way you think about people it's got its own checkered racist history as well of course yes and so i don't know i think moving towards i think one one useful concept that's being in a a lot of kind of theory of ai these days is generality um which i think uh so talking about artificial cognition and generality wouldn't i i think that could possibly be a good thing for the field if those kind of concepts became more entrance relative to intelligence because there's also this annoying paradox there is fairly strong there's this problem with

1:09:57 intelligence of the concept in um ai which is that you know once once a system can do it it's no longer intelligence right yeah exactly it's like chess is what intelligence is or the computer can do it that's not what we meant right exactly yeah i mean another recent one one one test i really liked and got totally suckered into oh this has got to be the test of intelligence is uh the winner grad schema challenge yeah so this is basically uh yeah um if you were identifying the reference pronouns so by for example that would be one case so if i say um

1:10:30 the um the scientists gave the monkeys bananas because they were satisfied with how the experiment was going you know that the second day refers to the scientists if you say if i say you know scientists gave the monkeys bananas because they were hungry you know that they refer to the monkeys right but obviously there's nothing in the sort of strict grammar of english that tells you what it has to be it's just the fact that you know that it's background knowledge that gets brought to bear enables you to answer those satisfactorily um but recent language models so they i mean this gave rise to a whole winograd schema challenge and um the winning grad schemer is uh part of a variety of

1:11:05 benchmarks like glue and super glue for natural language understanding um but yeah i mean artificial systems using basically fairly shallow statistical heuristics can do really well on this stuff in a way that caught me by surprise i thought here we finally have something that requires common knowledge common sense um but no again statistic can go a long way with statistics um so because i would the original question i was asking it was like how how much do we need to build into these systems before

1:11:37 uh i mean to count to have them count yeah so this was more about the moral patiency stuff yeah um so so when when does when do we have to start worrying about the roombas and self-driving cars of the world um i know you have a whole paper on this so i wonder if i could just get you to talk a little bit about that and i want to connect it back up to the thing we were talking about before about these like the folk psychological notions about you know uh whatever so anyway can i get you to talk about the moral patience yeah absolutely um let me just sort of preface this by saying that i think the psychological grounds moral patient to me is one of the most

1:12:10 important topics in philosophy of mind these days and it's uh relatively few people working on it and um i mean you and i are both you know into animal rights uh i think um the fact that we have not had a conservative scientific philosophical effort to figure out sort of our appropriate obligations different kinds of living things it's like such an oversight it seems to me and i think developing a really trying trying to move towards a more scientific basis for us thinking about animal welfare is really important i should clarify

1:12:43 here of course there are people particularly in like more applied animal welfare sciences who are doing really good work in this area it's the sort of the philosophers and queens of sciences i think could be doing more to figure out you know how do we evaluate the kinds of states that are morally significant kind of beings that are received so i think artificial moral patiency um can be seen as part of a broader project that i think would be a really good goal of trying to establish uh what kinds of beings matter at least matter in this special sense we associate with psychological patients yeah you might think you know beautiful

1:13:17 works of art matter beautiful mountains matter but there's a special sense in which you know beings that can suffer for example matter and trying to establish which beings qualify as uh as moral patients in that sense seems to be a really important project for philosophy of going to science and of course by the way awesome yeah and it's funny by the way again this is probably something we agree on but i it's so what um you know so much of my work even both on this and in other areas they are uh and uh there is a philosophy it seems to me that people just don't think about animals very much at all

1:13:50 um there's this notion in uh uh robert anton wilson's uh illuminatus trilogy of these things called florence i'm not sure you come across this but there are basically things that people are programmed not to notice and um it seems to me that we are i mean obviously people love their dogs and cats but there's so much of um animal suffering and also just the fact that we share the planet with millions of species of beings that all have their own kind of cognitive cognitive cognitive frameworks and ways of thinking about the world approaching the world is something that we go through our lives mostly blissfully ignorant of

1:14:22 and um to draw another sort of connection here uh i think like a lot of classicists okay but not a moral equivalent i should stress but i think a lot of uh classicists are kind of shocked that there wasn't among these amazing greek philosophers there wasn't a more robust serious discussion of slavery in as a practice in the ancient world um obviously you have people like seneca who have their thoughts and aristotle's an apologist um i think looking back the asian eight ancients you're like these guys had a real blind spot for this morally critical issue and i do think that in future generations we'll look back and say we had a real blind spot for animal

1:14:56 welfare issues anyway i'm going to get off my soapbox now but um soapbox appreciate it absolutely um so yeah i see the project of artificial water patiency as being really similar in certain ways and not really similar basically the same project as uh developing a more scientific account of uh animal moral patiency but the thing that makes it really hard is of course consciousness because most laws what most people see is the possible views of moral patiency at least in the kind of really heavy duty sense appeals to something like suffering and suffering most people seem to think only makes sense if we're talking about

1:15:29 conscious beings or conscious experiences um and there's a you know to give a very simple toy argument for this you know the fact that you're uh you may be getting tons of not susceptible input when you're under the knife and you're a general anesthetic that lost deceptive impact is morally insignificant because you're under a general anesthetic assuming by the way i know general anesthetics of themselves would be closer but assuming you really are unconscious right exactly any nonsense so there's a really intuitive idea that you know um negatively valid states only have moral

1:16:02 significance the extent of their conscience but if we yoke the project of trying to establish criteria from all patients seeing animals and ai to solving the problem of consciousness then that's terrible situation to me yeah good luck with that exactly right so um i think one important thing we can be doing is maybe moving towards fuzzier vaga but still at least partially scientifically informed heuristics thinking about ai more patiency and moral status in

1:16:34 animals so i mean the one phrase i use i just really like the phrase in a paper on ai while patients is an ecumenical heuristic roughly the idea is to get something that people with very different theoretical commitments can agree well yeah this gives us kind of this is this has some evidential states when thinking about the moral status of uh the given ai system we've got this metric this ecumenical heuristic that we can use that will cause us to increase or decrease our prior probability that this uh this probability that a given system is conscious i think moving towards those

1:17:07 kind of convergent measures is something valuable um we can be doing um so in the uh you know in the paper on um i mean uh uh paper on ai moral patience i suggest um building on workout john danah has i think i suggest there's um something it's just basically a cognitive and idea of cognitive equivalence so to the extent that uh an ai system um can be robustly described in the psychological capacities in terms of the same psychological capacities

1:17:39 uh saying the same relevant psychological capacities yeah as an animal to which we already attribute some moral status then we should attribute some some corresponding moral status to that ai system now that is incredibly weak this is just uh the weakest possible heuristic but yeah the core idea is if if you've got an ai that is in the relevant sense psychologically similar to a dog and we care about dogs then we should probably care about the ai too of course the devil is in the uh phrase relevantly psychologically similar

1:18:12 um i mean i i think that's tough but i do think there is it's not just the total handle 80 clause i think we can get it we have both intuitive pre-theoretical ways and some scientific ways of trying to figure out which capacities do you seem to matter and which don't um you know so capacity to um in the case of in case of uh maybe maybe doing certain kinds of navigational tasks right it doesn't seem like there's a powerful intuitive connection there to anything associated with moral status um so that kind of cognitive metric uh roomba's out groom is out exactly

1:18:49 um i mean i think my main worry is that um about this field is that uh we are going to go into um this age of increasingly smart machines um without and while still mined in philosophical controversies without any real good guidelines even more guidelines to sort of help us avoid ai suffering and also just in case any listeners are thinking you know any suffering crazy idea uh i actually think the important i mean we are likely it seems to me to

1:19:21 create ai's capable of uh of suffering long before we hit something like agi artificial you know human level and artificial intelligence um i mean i based that it's a very simple sort of argument but i mean i humbly most of us would happily attribute the capacity to suffering to a pretty wide range of biological creatures you know i mean there are really interesting debates at the moment about whether insects feel pain for example um and so i think it's entirely plausible i mean not sorry it's entirely possible because there's a huge amount of uncertainty we don't know what kind of architecture is going to be a mainstream

1:19:53 but a really significant worrying moral possibility that we'll build architectures um uh that don't command our maybe in some cases don't command on most because they're just maybe you seem like these really complex databases for example yet nonetheless instantiate the kind of core gears of the requirements for some kind of conscious capacity for suffering um i think that's a real risk and we should have yeah some criteria avoid that i should also add like a second please no add your addi edition uh yeah the adding thing here is i should also

1:20:27 mention um and i guess this is kind of core consciousness live talk your view on this issue might partly come down to how optimistic you are about the current state of the science of consciousness so i think um maybe ten tens more recently five or six years ago even it seemed to me that we were really making very rapid uh not just progress but convergence in the science of consciousness you know it seemed to me that there were a few really big theories and um the differences between them were in some ways getting made more salient or in other cases being kind of uh being differences were

1:21:01 being divided were being bridged and you know we had obviously things like workspace theory thought theory but that we were converging towards some kind of consensus of progress yeah and over the last sort of three or four years and this is purely my impressionistic take of uh going to a lot of assc meetings i feel that that progress has uh if not reversed and it's sort of stalled there's not the same kind of momentum okay so it's not just me um and i mean you know like suddenly uh views that i sort of thought were kind of marginal like psychism uh

1:21:36 back and back in there back in the topics of discussion like maybe they absolutely should be right but i mean well that does suggest to me is the idea that we're moving towards a point of scientific consensus about the nature of consciousness i've lost faith in that to some extent i mean maybe in over the years that they look totally different but to the extent that timelines matter here's the extent we are we think there's a good chance that we're going to be building animal level and fairly high level animal level intelligence ais in the next 10 to 15 years you know we need to have something in place even if the science of consciousness doesn't give us the results we're looking for yeah no i couldn't agree more with that

1:22:09 the possibility of creating suffering that we're sort of um not not sensitive to is is something that should be taken quite seriously and i mean so i like your approach right because you kind of build on this theory well i guess you you reject the theory light approach right the theory light approaches yeah so i should say a little bit because i think this is another really interesting issue and again it's one of these topics um where it seems to me again so much of the interesting the areas of glossary i'm interested in questions about animals and questions about ais collapse into a single interesting set of questions about non-human minds

1:22:43 um so a recent paper of mine is forthcoming in my language um on the specificity problem as i call it is basically about the challenges involved in attributing consciousness and applying our existing science of consciousness to uh non-human systems because you're quite quite by the way this is not an original argument versions of it have been going around for a while and you know eric schwartzgaber justin prince and peter brothers have all sort of given versions of this but in this paper i kind of try and wrap it up nicely with a bow and say here's the problem and basically um

1:23:15 here's one way you might think we can settle questions about whether say insects are conscious you might think okay we take our current off-the-shelf best theory of consciousness maybe that's hydraulic thought theory maybe that's called the workspace theory and we see whether the relevant system instantiates that theory whether this theory is true of this system so that's this is where you run into the specificity problem and by which i mean knowing how fine-grained how specific uh the version of the theory that you take from humans and apply to uh to non-humans has to be so there's a whole bunch of capacities

1:23:48 that uh we associate with workspec the global workspace you know standard harm talks about capacity for verbal reports as a criterial feature of availability of a global broadcast and you know we've got tons of stuff we can do with globally broadcast information memory etc um which of those are essential and which of those are non-essential i mean so another way to ask that is to say how i think this is eric schwabel's terminology uh how similar to our global workspace does some kind of broadcasting system in

1:24:21 a non-human creature have to be in order to count as a global workspace right this is a snails problem right they have something like a global workspace but not all the like grandiosiosity of it or whatever so does it good is it good enough absolutely yeah exactly um and i think i mean higher thought theory here is interesting i think higher effort theory maybe has some better answers to this than global workspace theory but even then i think you might think look particularly if you have a messy view of the mind which i think i'm sort of sympathetically inclined to where you get lots of stuff that sort of uh

1:24:53 doesn't easily fit into into neat categories like conceptual versus non-conceptual for example um uh you might start to worry like how how fault-like does some system of representations have to be in order for it to count as uh high water thoughts in the relative sense and i'm not saying by the way this these questions don't have answers um no totally yeah they're just really tricky to answer um so i think one of one one way around this um if you think this is this is a nightmare there's been we kind of apply our theories of consciousness

1:25:26 to um to non-human systems um jonathan birch has got this really interesting uh um view out of paper he just published forthcoming news i believe on what he calls the theory light approach which roughly is you take um behaviors or capacities that seem to be very reliably associated with consciousness in human beings so classical example the one that i'm sure some of you listen to familiar with is uh delayed eyeballing conditioning so you know the idea is if i uh i flash this thingy stimulus to you um threshold so you can't see it

1:26:00 um and then it's immediately followed by a buffer there you will learn to blink your eye even though you don't you're not aware of ever haven't seen the stimulus but if you interpose a delay of a certain threshold then you only learn that reflex if the stimulus is uh super threshold okay right this seems to be kind of learning capacity that we only get when stimuli are perceived consciously um and uh jonathan birch discusses a few other similar cases certain forms of reversal learning for example some kind of multi-sensory learning the ability to

1:26:32 transfer knowledge acquiring more modality a play in another and in humans they seem to require consciousness or seem to be reliably associated with consciousness so his idea and i think again he would say this is sort of again at the level of building an evidential case it's not you know hard to the extent that we find um uh find that these capacities are present in um in creatures uh in non-human creatures that is good evidence for these creatures being conscious in other words put it very simply if there's something we can only do consciously and we find another

1:27:05 creature that can do that thing that's some evidence that that creature is conscious now obviously there's a lot more to that you might think maybe there are background conditions that are present in humans that might be not present in these non-human systems such that sure when we do this it has to be conscious but because the background condition isn't present in these non-human creatures when just when they do it doesn't tell us anything about consciousness at all so i mean there's a lot of nuances but i think it does what i think is critical is it maybe provides a relatively yeah his theory light way to get into the problem maybe gives

1:27:37 us another hook and um my only sort of worry about it is that uh depending on particularly when we want to apply this not to relatively phylogenetically close organisms like birds or even insects right but to systems like ais have radically different architectures from us potentially the odds that these uh they'll display these behaviors uh going to be significantly lower right right i mean i mean and it seems the fact that i mean and i'm sure in many cases um so you've got kind of a double worry here about false positives

1:28:10 and false negatives so maybe you build an ai and it's can't do reversal learning it can't do the late idling conditioning and it can't do multi-sensory learning that's just because it has a different architecture it could have been incredibly smart having rich conscious thoughts but just because it you know it's not built like us it's not gonna um it's not gonna be able to have these specific capacities and if this was the only test we were relying on we would get a um a false negative one but there's also the kind of false net false positive worry right which is right if you were to make these criteria then you know give a uh give a smart group of computer scientists uh a bit of money in

1:28:44 a couple of weeks and nothing they'll come up with a system that does it by the way i think it's a more general worry for a lot of theories of consciousness you know um this was one of the first things that i uh i ran into speaking to more computer scientists at uh cfi it's like i said oh i think there's this theory of consciousness oh i can build a system we can do that in a weekend so well maybe you can but there's also this theory of consciousness oh yeah that's easy too um this kind of a trivial realization yeah exactly i mean so just to sum up some of the stuff you're saying so the problem is that

1:29:16 you know there may be systems that are like these huge databases um and uh search engines or something like that that are very complex and have all sorts of interesting things going on but no similarity to anything that we recognize and so how can we say so it's one thing to say you know oh a beetle versus a dog that's already hard people already have you know different views oh the beetles nothing dogs everything but then when you have something that's not in any way like in any animal we've ever encountered how do we transfer our uh

1:29:51 our moral thinking to those sorts of things we're not ready for that i mean i just don't think that humans are in any way up to this challenge i mean i worry about it a lot and uh you know i even sometimes worry about you know the judgment of the wider universe i'm gonna sound really space-age now right but i mean i don't know i think uh if there are people if there are other civilizations out there that are eventually gonna get to meet and if we've spent the last couple of centuries inflicting massive suffering on artificial beings and maybe they are maybe they'll be artificial beings themselves

1:30:22 it's not a good look for humanity right no i mean as soon as they learn about factory farming it's over for us but yeah right if you had to if you had that other stuff it's like not quite good um but so what's uh what's the way out of this like what's the best way out of this on your view well i mean it's not a way out entirely but i mean what i suggest in that paper specificity problem is that maybe we can marry the two approaches and the details are a little bit messy about to sort of me give the kind of broad view the idea is that theory-like approaches give us some independent these kind of markers approach their signatures of

1:30:54 consciousness approach and not not some neural symmetry like you know but behavioral signatures things that you can only do that humans can only do unconscious um they give us some kind of independent hook some evidential hook as um into the problem so let's just say that you are trying to apply the global workspace theory to an animal an organism say a lobster and it turns out it can it meets a whole and you cast it at a certain level of specific say i'm going to include this in the theory not include that and then you're trying to apply it to a

1:31:26 lobster and uh clearly the obstacle clearly doesn't satisfy this but according to the theory-like approach the lobster satisfies all the criteria right it can do all the stuff it can do delayed idling conditioning it can do reversal learning then you might start to think oh maybe that's the wrong level of specificity um to adopt it in the first place maybe we should try and engage in kind of dynamic equilibrium between on the one hand these markers of consciousness approach and our theories and we can use the the one to constrain the other i mean it sounds like there's a risk of

1:31:59 sort of already taking our background assumptions about which things are conscious and which things aren't yeah as as we used part of the fine-tuning yeah i mean i think that is a that word i mean there is um i mean it's not just that right it's not just the case of uh uh well clearly or just conscious right i mean we could be surprised we could find that these markers of consciousness these things that humans can only do consciously uh the you know creature that we don't we wouldn't intuitively think of as conscious investment maybe an emotional worm doesn't right um and i think that so the theory light approach isn't just you know bringing

1:32:33 your predecessor intuition to the bear but i mean i do agree with the intuition problem on the other hand this is a gnarly problem that i i run into i mean i i think again there should scale that snail paper puts it really well but we're more sure that dogs are conscious than we are that any particular theory of uh consciousness is true and um i mean maybe not everyone shares shared that people there yeah i mean i would just say though that dogs are special uh given their evolutionary history and close contact with humans they develop psychological i mean

1:33:06 they're the only animals that care like whether we're happy or not they're one of the few ones that like notice when we point at something or what where we're looking so they do have some psychological traits which are are unique because of their close association with humans but i agree about i agree about the dog thing right just we uh but i mean the so so anyway you could go ahead you i have something i was gonna ask why go ahead yeah no i mean um so i mean i think it's really hard to try and figure out how much evidential value to place on our pre-theoretical ideas about which

1:33:39 systems are and are not plausible consciousness candidates clearly those intuitions aren't perfect you know fluffy animals with forward-facing eyes obviously conscious right really that's not something we could care about um but i mean i i feel that i guess we would both confidently save you know any theory of consciousness that said uh children aren't conscious until the age of 16 is obviously false we can [Laughter] [Laughter]

1:34:16 um yeah i uh so i think there's some evidential value that should be placed on these theoretical intuitions i mean and it's it's kind of tricky to figure out exactly how to cash out why that should have any epistemic status the brand abroad strategy i'd say is something like our concept you know there's this talk about this line from klein famously where he he's accused of sort of uh renouncing the line of two dogmas where he says he won't worry about changing the subject yeah i forget exactly where he says it but um but i mean there is a worry right that if the kind of attributions of consciousness that a theory of non-human

1:34:50 consciousness started to make started to get too wacky there'd be this worry that it it changed the subject basically but it was no longer tracking the concept we started with and at least some making some contact allow free theoretical commitments about the kinds of systems or the kind of behavior that are indicative of consciousness some such kind of context is required from theory you know to be about the thing we wanted to be about yeah but i mean i think it gives a that's the best way i can wrap up the fact that we i don't want to ditch these commitments entirely i should clarify that here though that pan psychism is an obvious worry for this um and i don't know i i was a panzers for a

1:35:25 year as an undergrad i'd actually say a grad student i think it's a really cool view and there are some problems that it tackles really well um and i think it's a big i mean it's a big worry though because i've had two hand psychists well-known ones actually tell me that the reason that they're not sympathetic to vegetarianism or veganism is because pancychism basically that everything that you know everything suffers and some or everything is conscious you've got to eat something two people who if i won't reveal their names but if i told you you'd know them that you know these people so this is i

1:35:58 do think that's a worry is that if these if their view this abstract theory of consciousness i mean first of all it's kind of silly because obviously the difference between the chicken's mental state and the mental state and electron so you know one we should worry about more than the other obviously even no matter what your views are but i do worry that your your antecedent views about consciousness can have these kinds of impacts on your moral thinking yeah that's why i was gonna i mean i was gonna ask you why so not to interrupt or cut out but so because my own point of view you know i'm i'm i was raised vegetarian i'm a vegan i've never had any meat um i'm pro

1:36:30 animals but i do think there's a methodological issue about determining which animals are conscious and i think that is really challenging so my own kind of resolution of this is to say look real science of consciousness should be done on humans um and then once we figure that out we can talk about animals but i also think that so what do you say about you know the cows and chickens if you don't want to say the science of consciousness on them to me there's there's a moral argument there like if there's the ben if there's a question then they deserve the benefit of the doubt um and this is like an argument like jeff sebo i don't know if you know him

1:37:02 or not but he's made a similar kind of argument with respect to insects like if there's even a question that they're suffering and if you're kind of utilitarian you think like you know if they suffer even just a tiny bit in the millions of ants it's going to add up to a lot of suffering so we should give them the benefit of the doubt there and that's not an empirical claim that's a moral claim yeah and so one wonders like why why isn't that uh a way to you know that these aren't empirical questions these are moral issues and then when we we can deal with the empirical questions but isn't it dangerous and i think you know maybe you know someone i think singer says this in his animal

1:37:33 liberation book it's always dangerous to pin your like moral ideology on some empirical question because it may turn out in a way that you don't like yeah oh so so many cool issues there okay so um so one i would say that well you say tomorrow question modern empirical question but i guess it's still going to be a bit of both right i mean even if you adopt a fairly aggressive precautionary principle there are going to be some things that aren't covered by it right you know the probability that rocks are conscious is going to be sufficiently low we don't need to worry about rock welfare and so how how widely the proportionary principle extends is going to be constrained by empirical

1:38:05 factors right uh which kind of creatures are science can give us clues about which kind of breaches are made vaguely plausible consciousness candidates and i don't know if you'd want to uh endorse this but it seems to me the best way to apply a precautionary principle would be to sort of roughly take some probabilistic uh calculation of the likelihood that the likelihood according to the current science the common system is conscious and then maybe add a precautionary multiplier to that in terms of doing your sort of decision theory about what you can and can't do um but i mean there's also this really

1:38:39 related this this related problem it's something i've been thinking about a ton recently and i've still got a couple of draft papers where i try and wrangle with this um about intersubjective welfare comparisons and into subjective moral comparisons more broadly um so you talk about sort of from a unique utilitarian perspective what about the millions of ants um and so one argument you hear from effective altruists is that if you're going to eat one animal you shouldn't get any animals but if you are going to eat animals better to eat beef or chicken right because um

1:39:10 a cow and a chicken in through the process of factory farming you're going to suffer a vaguely similar amount but at least with a cow you can get a ton more dinners out of it than you can out of a chicken right you know the logic are a lot more intelligent wow i mean right okay i mean i was deliberately kind of controlling that but not um and then assuming in two animals of equivalent intelligence you should choose the one that um sorry not because intelligence two animals that to which you assign a relatively similar probability that they're conscious right uh you should eat the bigger one right so and if you're a doctor i'm i'm

1:39:44 sympathetic to a very broad view of consciousness in nature i think probably consciousness may my views here are very subject to empirical revision um but i've been impressed by arguments like by videographer smith of course but also um feinberg and mallet uh a really great work on the evolution of consciousness wonka and uh simon ginsberg have a great book as well on this thing so the the kind of the idea that consciousness has a pretty deep evolutionary origin possibly coinciding with the development of you know at least slightly fancier central nervous systems so if you take a pretty expansive view of consciousness

1:40:16 then you know to give it to be very crude about it you should be eating whale meat right ecological consideration inside right you know one whale i mean and whereas you know if you order the spaghetti alcanbary with like 30 prawns in the dish that's like a that's a devastating moral tragedy right but i i also kind of do worry about there's uh this can kind of lead us into absurdity um yeah it seems yeah going back to these sort of i'm more confident that of this pre-theoretical idea than the truth of any given theory um it seems to

1:40:50 me that saying oh well look these ants can plausibly suffer so we should kill a human child to save 10 ounce it seems to me that's just i'm more confident that that's wrong than i am in the truth of whatever clever theoretical footwork got us to that place um now of course you can say oh yeah but you know you wouldn't have to endorse a view like that because you know the ants even if they're conscious and suffering they count less but why this is what i it's something i've really been banging my head up a head up against obviously there's the famous ben from quote the question is not can they reason etc can

1:41:23 they suffer yeah that's in a footnote um but i mean if you think suffering is like really what anchors our moral commitments um then and you think consciousness and suffering are pretty much ubiquitous in a wide range of animals then um i mean i think you do get led to some pretty extreme results unless you can provide a non-ad-hoc non-self-serving reason for thinking that humans should count a little more i mean i thought the traditional kind of utilitarian answer

1:41:56 was that they're capable of higher levels of suffering so you know they can anticipate their own death they can be you know they can they're afraid of what's going to happen to them while they're being held captive but that doesn't seem plausible for hands and so forth yeah i mean i think that definitely gives humans a bit of an extra edge um but how much of an edge it gives i don't i'm not i'm not sure i mean so just to give an example i mean if you ask me about the worst experiences of my life right um i think probably the one that i would easily singled out is a horrible case of

1:42:29 food poisoning i had where i was throwing up in front of the toilet dry heaving for three days straight it's like oh you know like i mean i've had more significant bad things happen to me i obviously you know grief is something we've all experienced but in terms of raw awfulness i think many of the most awful things we experience take the form of very simple sensory awful sensory states and i don't see any reason to assume you know absent some good yeah any and get through through every reason to assume that the client if we grant that ants can feel pain and that's can suffer that

1:43:02 what it is for an ad to be agonizing pain is any less smartly significant than what it's like for me to be an agonizing man right and i think sure you know humans might get a bit more of the agonizing pain because we can you know we can uh expect it in advance but you get enough ants on the table and humans are going to be it's not going to take that many ads in order to accumulate accumulated experiences with the interests of humans i mean look at some point right i would absolutely have to agree that you know if we're talking about humans versus the biosphere then you know us right um

1:43:36 you heard it folks henry several and uh humans hopefully when it comes to that yeah exactly um well i am on record as arguing that we should work towards uh photosynthesism so i i actually think this is kind of puts the problem of evil in a whole new perspective for the you know prompt for god uh why aren't we photosynthetic uh it is possible but anyway so yeah i think we should abolish you know abolish all forms of carnivorous activity even eating plants i'm not sure why uh

1:44:10 why things turned out that way but um yeah so i i do in a perfect world i would opt for photosynthesis yeah because i think these i mean i think these are thorny issues you know so you know you have a hundred thousand ants versus a human okay obviously i prefer the human but there's isn't really a good at i mean if you're committed to the maybe you just say ants can't feel anything i think that's you could argue that view but if you're committed at all or if you want to give them the benefit of the doubt because they're such complex creatures um then you're going to work yourself in all kinds of pickles and we don't have really good

1:44:44 i mean we're bad at moral theorizing as a species anyway but in this area this is one of those places where we're extremely bad which is why we need some ai to help us with these things yeah that would be nice moral art yeah i mean go ahead though guys quickly just wrapping up on sort of the vegetarianism angle maybe because uh you know i'm a vegetarian uh and you know we both care passionately about animal welfare um but i think the one sort of argument against vegetarianism i've heard i'm curious what your view on it is this

1:45:15 this familiar argument right which is that well assuming we could get rid of factory farming and adopt modes of animal pastural farming where animals lived on balance good lives it would be better to live in a world in which we have lots of animals where we have more happy animals around net happy animals as a result of carnivorous practices remaining in place than a world in which uh we ditch carnivorous practices but as a result there are just parkier animals around yeah i'm i i don't yeah that kind of argument is like

1:45:49 uh i don't know how i should put it but i find it very sort of morally affronting so the average cow like killed at two years old that 10 of their lifespan so that would be equivalent to killing a human being when they're like you know 1920 something like that and saying okay well you lived a good enough life so you know you're not entitled to the rest of it so already right off the bat uh it just seems like well so unless there is a really significant need for the protein um i'm not sure what in the world could justify that's certainly not the taste and i understand that it's a complex

1:46:22 world when i was younger i saw things very black and white but i understand there are people that can't afford that don't have access to other forms of food this is what they have to do so if you but if you factor that out and all things else are equal um then i just don't know why we need to justify why we are looking so hard to justify the the urge to eat meat i just don't understand it but of course i was raised vegetarian so for me i would rather eat this table than eat an animal it just doesn't make sense to me i never thought of it as food i was raised to think of animals as friends as companions so it's just that it's very hard for me to work myself into that

1:46:55 point of view but yeah i could see why it's not as bad as what we're doing but that doesn't make it good so i i just don't know yeah i mean i think the the most sort of the way that argument i've usually heard it um just realistically given people's priorities right a world in which everyone is vegetarian is a world with dramatically pure animals with vastly more land use reserved exclusively for human activity and okay given the present shitty state of fact farming that world might still be better than uh this world but maybe a third world that would be better would

1:47:28 be one in which we keep animals as part of our lives which again ex-hypothesis we're only going to do if we keep on eating them but we just make sure they're well thorough taken care of i mean yeah but so the reason there'd be less animals the reason there'd be less animals is because we wouldn't be farming them so i don't think it's a heart so i don't think it's a harm not to be brought into existence for any future cows that don't get born um so i don't think you're harming them in any way so there would still be plenty of i mean there was this recent paper about getting rid of carnivores like if we could breed them out i don't

1:48:00 know if we read that or not but i'm for that i think yeah why shouldn't we genetically engineer what gentleman i know he had to call him in yeah exactly so i'm for that it's like yeah we'd get rid of a genetically engineered tiger so they don't need meat i'm 100 for that um i mean a world in which we i don't know we were so appalled by the ubiquity and inevitability of suffering in animals and so we decided to um painlessly euthanize all non-human life all on human conscious non-human life you know maybe we keep gene banks and these things and we continue to gain

1:48:33 scientific insights and we replace you know the ones that we we replace them with totally non-conscious animatronic replica so people can still you know enjoy seeing the birds fly in the sky that would be a better world than our world um no because you'd i'd say you're committing you're doing something wrong by euthanizing them um but uh but if you just let them all if you sterilize them all but and then not let them and let them die off natural lives and they never reproduced i would say

1:49:04 uh they would be bet they wouldn't be better off because we would be impoverished with respect to this but i don't you know i think think that would be less bad than the world you were talking about i see so you're pretty sympathetic to some form of negative utilitarianism maybe well i'm not really sympathetic to utilitarianism in general i'm kind of a continent in this area but uh basically i think you know i think the categorical comparative has some intelligence about animals and i know you know christian coastguard has written papers on this that i find very uh influential but even before i read those papers i sort of thought the

1:49:36 categorical imperative um has some something to say about you know uh if i you know you know just universalizing a principle about treating animals in the way that we do is not something that uh not maybe a rational being would submit to but a sentient being um right ascentia beings ourselves i don't i see a kind of inconsistency in that attitude there but uh i'm all but in so far as i would be in a utilitarian you got a negative utilitarian yeah definitely because i mean um i guess i'm maybe more sympathetic i mean i'm not eating tyrion

1:50:09 but i'm maybe more sympathetic to at least something comparing arguments um i don't mean this again this is a bit zany but i think one more optimistic way of viewing humanity is that we are the vector by which the good that is the human the global the earth's biosphere the other vector by which that can actually spread throughout the otherwise inanimate universe or maybe not venus venus has already got who knows you know all these worlds that have no life on them that maybe if everything goes well

1:50:40 um we will enable you know pretty quickly they're going to be no dogs on mars unless we put them there yeah the the kind of rich potential future biospheres um standing across maybe thousands of otherwise barren worlds that humanity could engineer for the earth's biosphere that could be our sort of um that could be the thing that justifies humanity's existence i mean that's a nice thought as long as we yeah so i mean i've i have a antagonistic relationship with nature so i find nature terrifying um and not very friendly it's beautiful

1:51:13 obviously and i enjoy nature but i mean just the the sheer brutality of nature is off-putting to put us mildly but uh that's okay say what a herzog fan it makes me think you know that's a theme yeah yeah concerning that exactly exactly yes um so if we did that and we spent more we could be spreading more suffering uh so we just you know if we create a lot of carnivores uh and transport all the tigers and antelopes over to mars and i don't think that's a net good overall um but uh certainly i think that's one so i

1:51:46 you know i i made this joke i almost wrote a paper on this but i made this joke a while back that you know it would be nice if you know all the turkeys that are killed on thanksgiving if they first uploaded them to a virtual environment so the turkish could go on like living and then they killed them and ate their bodies but then they would be in this virtual like barnyard basically living forever um and you know i i that was mostly a joke but when you think about it actually if it's possible to upload consciousness into a computer at all then it's it's not it's more than a joke it might be something that we should worry about uh

1:52:18 of course i mean i don't know so i mean we're running out of time here and i can't keep you forever but i do wonder about your views about uploading yeah um yeah uh uh poorly speaking i think um hell i'd take the option um it would it was the choice between you know if i if it was choice i mean obviously you know if it's if it's if there's zero cost i'm gonna die and the last thing that can happen between dying and uh i got one my only church of surviving is uploading and that's a bad i think a lot of people would take just you know just just on the off chance but i mean i think even under more demanding versions

1:52:52 where it's you know you can live another 10 years in your biological form or potentially a thousand years been uploaded for potentially forever i think i would still i think i'm still pretty optimistic about my capacity about our capacity to both be conscious and retain some relevant sense of identity um i mean i'm very sympathetic to sort of poverty and views of uh personal identity according to itself it doesn't really exist in any kind of robust metaphysical sense and um i mean one way i think about this is uh this is this is brutal right but imagine

1:53:26 like someone you cared about like a spouse or even a child right had a terminal illness and you could upload them right i don't know people might have very different intuitions on this but i mean mine would be hell yeah i mean yeah even if it's not completely them even if it's uh there are some differences but even if it's not as much them as they were before at least you're giving them some kind of survival have you seen uh have you seen devs do you know this show okay this is this i want to spoil it for you okay okay cool but

1:53:57 but so never mind forget i said anything about that but so so if you do so this is the question i was going to ask you if you do think that uploading is a possibility that you'd be conscious and even some sense of continuity would be there then my intuition insofar as i have it because i'm kind of a skeptic about uploading but but i you know i don't rule it out i think it's a an empirical question like i said before but if it were possible then my intuition would be that it would be a great moral crime to upload all the animals in their current state um so that you know that they could live like you know uh

1:54:31 you know tigers hunting antelope and so forth that we could prevent that and not modify them anyway i think that would be it would be morally wrong for us to do that do you what do you think about that oh that's that's a really good question i mean i think maybe something my initial intuition is the same as yours i mean one thing that may be creeping in uh the influences our intuitions there might be that it seems like we have an easy opportunity there to reduce the amount of suffering right as we upload them um [Music] um and uh i think that might be slightly the

1:55:03 intuition so um i mean here's a crazy experiment something something that sounds like it's from a three-body problem or something but uh let's just say you know some terrible event was gonna wipe out the us biosphere and we could upload the biosphere into digital format but because the uploading process we couldn't tweak it at all so either the biosphere was going to be painlessly destroyed by some gamma ray burst or we could upload it but we couldn't fiddle with it at all right in that case that then my intuition is quite comfortable oh yeah upload it

1:55:34 you're saving a huge amount of just for the benefit of the organisms themselves the creatures themselves sure they're still going to be suffering but um it'll be it'll be better to do that than allow them to be wiped out by uh by this camera burst yeah um that's interesting so so you think that in the original sort of case then the because i do agree that there's a sense that well maybe we could fiddle with it a little bit so that the animals don't have to hunt each other or something like that yeah but it's not clear to me that like if we did upload we would still need to eat

1:56:07 right meat or anything actually in the virtual environment i mean there would have to be something that gave us the energy uh in the virtual environment and if you know you buy the simulation hypothesis it could be just like this right so you could have all these experiences but it's not it's the scene in the matrix my cypher is tucking into the steak in the middle yeah exactly oh i know it's not real but it tastes just like the real thing exactly so if that were the case then yeah if the animals had to actually be killed and suffered then yeah i'm not sure i have the same into i have more i struggle with i would uh

1:56:39 yeah i have more of a pull towards let it all go you know i could why continue the great cycle of suffering what's so great about it i mean i guess um i mean that's interesting you use that phrase i think one of the maybe deep fault lines i think in ethics but it's not a deep fault line in terms of most debates i think i've noticed it seemed to be a bedrock moral principle that uh people disagree on it's between the kind of anti-natalist or more negative conceptions of uh of well-being and the good and

1:57:12 views that take life itself or conscious experience to be you know fundamental good and that the world is better it's dramatically better for having more conscious beings in it seems to me that's quite an interesting that's quite an interesting fault one um and i think we're probably on different sides from the fault line yeah a world in which humanity can spread the biosphere across billions of planets even if it creates more suffering is still a better universe than one in which you know we all quietly die when the sun uh suddenly becomes a red giant yeah i i guess i wonder why in what

1:57:45 sense is better yeah i mean and there's a lot of different moral iterations that interrogate i mean i'm i mean i think one interesting another source of interest uh something that's definitely heating into my intuitions here is uh i guess i have uh sort of an ecological some strong moral more ecological moral intuitions that even aside from consciousness and i do think consciousness gives a special kind of moral value but i'm inclined to think there is something just valuable about life and the biosphere in itself that yeah a world full of non-conscious um

1:58:20 beings um conscious complex living systems is still a better world than a totally barren world um yeah so but so a world of zombies is better than oh or a world of plants which one are you talking about i'm thinking more i mean thinking of the planet's case specifically right um but in other words but i mean the zombies case works as well the value of the value of life isn't exhausted by the value of consciousness even if consciousness is sort of the the mother lode where most of the value resides yeah actually it's actually quite hard to justify that notion actually that consciousness is not that is valuable at

1:58:54 all or gives or matters um it's very i mean other than you know thus just pointing to your experience clearly you know um i mean are you talking about consciousness per se or you also you're not even like suffering no just consciousness per se right yeah you know david chalmers has this thought experiment where uh he considers the race of vulcans they have conscious experiences they have perceptions um but they they don't suffer it's just all rational stuff mostly uh that's going on with them and but but they're conscious um

1:59:26 but uh and so he has kind of various trolley problems where you know there's some vulcans on one side and people on the other side and these things it would be wrong to like kill the vulcans um willy-nilly that they have some there's some moral value associated just with the fact that they're conscious even if they can't suffer yeah and i i see i i feel the pull towards that intuition to be honest yeah i mean my my take on that i think it's a really interesting poetics but my take on it is basically that um suffering being being with the capability for positively and negatively

1:59:59 valued conscious states suffering and pleasure uh is one way one critical way of having interests that's not the only way of having interests right and another way of having interest is being a reflective being who you know hopes that um your kid's going to get into a good school and you know presumably at least on the version of the first time i've heard these falcons can still have desires and uh can still think about things states they would prefer to happen things they prefer not to happen um i think that's another pathway to having interests um and so maybe i think the notion of

2:00:31 interests is fundamental and that suffering is one clearly plausible way of having interests being given but another way is through being sophisticated reflective uh being capable of you know forming goals and plans and projects for the future and then right but and you don't need consciousness for that i mean you could a zombie could do that yeah although you might think uh i mean it's depending on how uh how much you wanted to build into it you could say look consciousness is required for having interests but sentience is one way of having conscious interests uh having conscious speech directed

2:01:04 projects is another way of having interests and thereby restricted to the conscious to um restrict moral status to the domain of conscious creatures without restricting it just ascension creatures i created the conscious and concern right but i mean that's not the only way you can go and as i say i mean given that i um i think there's a compelling case to be made that um non-conscious creatures have have interest too and those interests have some moral status and i think that that sort of sense is maybe what anchors um my conviction that you know even a even a biosphere um devoid of conscious

2:01:39 creatures would still have some value because those creatures still have biologically interest that could be spelled out biologically in some sense but so if you had like a ai system that was very advanced but didn't have consciousness or even an alien species you know have you read uh peter watson's book blind site no you're not the first person to recommend it it's a great book but anyway in that book there are these non-conscious aliens um they they're very advanced they space travel they you know they have all these goals and so forth but they have no conscious experience at all they're to complete zombies um

2:02:11 and my feeling is that if we ever met something like that we would have obligations to them whether they suffered or not and whether they had perceptions or not uh consciously the fact that they i mean have goals i mean you know anyway so i think that there are other things which ground our moral responsibility to them um so you agree okay so i thought you were saying something that kind of was the opposite of that no i mean so i was saying um one view i described i was just saying sort of if you wanted to avoid the uh the arguments challenge gives against what you call sanctionism by only being

2:02:43 capable of suffering matter you could include the circle of care to include i think conscious being that didn't suffer on the grounds that they could have still have conscious projects but my own inclinations that expand even beyond that and include beings with interests even if those interests aren't conscious but i still think there is and you know i think uh uh dave has said that you know he's worried that that seems a bit ad hoc it's a disjunctive theory but i think it can be grounded in this notion of interests and that's what gives it this kind of unifying character that makes it uh it's not quite as hot as it may sound

2:03:15 yeah exactly i mean the the problem is you expand the moral circle so large you end up with like the dilemma of jainism right you can't you have to stand you stand still can't meet you you can't move i mean i was struggling going back um to your example the panzerkiss who say oh i should probably be vegetarian because um because you know everything maybe electrons somewhere um i mean that's a shockingly bad take it seems to be uh interesting but one i would disagree with um uh but i mean i'm sorry this is the point of slightly slipping back to point one is make earlier on that i mean i don't

2:03:48 discount the truth about psychism um and uh but if i was ever convinced that uh han psycho's theory of consciousness was true then it would seem to me that what i'm interested in would cease to be consciousness that would just be the combination problem right right i'd say oh okay fantastic is true okay well everything i've been doing so far all my interests are still good but they're now just focused on that subset of the con all all the stuff all the stepping conscious that subset that has these kind of psychological properties that we associate with moral status so electrons are not going to have i

2:04:20 think the idea of electrons have more status is just uh yeah it's too far removed from any prize i have me to even really make my sense of so i mean if but if it's true that electrons are conscious then it seems to me the interesting work becomes shifts from you know the metaphysical basis of consciousness to the metaphysical basis of conscious agents in the kind of thick sense that we care about right i mean i guess so i know you have this other paper i mean we i'm really running out of time here but um you have this other paper about motivational the motivational account of felt sensory uh you know the pain or

2:04:54 whatever the the negative effect of it yeah and so there you might want i mean so people some people have said things like well you know these things are repelled from each other that's kind of a like quasi-motivational state uh right isn't it i mean you can there's some so my point is there's some connection to our common sense notions about these sorts of things they're far removed but not completely removed so i mean i know that goth has said uh something similar when i asked him about um you know okay so if all qualia uh if all phenomenology is

2:05:27 you know basically built out of fundamental building blocks that are present everywhere then uh how do you account for phenomenology that has this um imperative character or uh this that feels good or bad that seems to be a fundamentally different type of phenomenology from say sensory phenomenology but it doesn't but it also seems to be the kind of phenomenology that uh we associate with certain kinds of psychological capacities having goals being an agent right right it seems hard to think how that could apply to electrons and he suggested he gave that exact same answer that you know attraction and repulsion is uh yeah that's kind of like motivation

2:06:00 i mean all i i guess there there's you know i'm sure we have time to go into the full details but broadly speaking i think it's a it's something really interesting about um the mind that we have this particular suite of uh mental states that have this quality that they can feel bad um i think it's a striking that a lot of states can't feel bad i don't think thoughts themselves for example can feel bad i think they often trigger negative emotions or negative sensations um so i think this this um it's a property of century and

2:06:33 somatosensory effective states that they have this capacity to feel good or bad and broadly speaking and being skipping over some details for speed i think basically what that suggests to me is that we have there is a deep reward system um rewarding punishment system that um is probably quite phylogenetically ancient another little clue that that's the case uh it seems to me is that um it seems like we can compare in a in a very imprecise way the felt badness of states in quite different potential modalities so you know you might think give a crude

2:07:06 example you know you're really exhausted you're lying in bed you wake up in another night the bed it's freezing cold outside of the bed but you really need to go to the bathroom is this worth it right and at some point it is worth it it's got to be yeah i mean i think we do that kind of thing all the time you know like we undergo we go through painful states because they will relieve our nausea or whatever um and so it seems and obviously you know we're not doing some clear adonis calculus in our heads we do that but it seems there is like an in principle commensurability of these very different states uh in nonetheless how these qualities of

2:07:39 foul goodness and felt badness which again seems to be clue seems like a suggestion to me that what we're experiencing then is the outputs of some quite deep reward system that's basically about the prioritization of certain kinds of behavior over others um and then you then i think that can also maybe even give you some clues into sort of the range of creatures that can have these kinds of states when you think about what kind of functions are these states sharing how many what extent do other non-human organisms share relevantly similar architecture to us and i see a peeker that's luke right that is luke yeah yeah

2:08:14 it's got a new baby sister as well yeah congrats on that by the way um so anyway yeah so i think um broadly speaking so i think there's a lot more to the kinds of states that we associate with suffering than just this true push and pull thing right it's a particular kind of reward system and i think desires are another case i mean i think desires are often associated with suffering or then or they're not fulfillment of desires but they have a very different character so simple example here is um basically i think states that can feel bad

2:08:48 um they provide a certain motivation to get rid of the states through any means possible whereas desires typically aim not just their own denial but a very specific pathway to their own removal that's a fancy way of putting let me give a really simple example if i want a cup of coffee and i go to the coffee shop and the guy says oh we don't have any coffee but i got a pill that will make you stop wanting coffee but if i've got a pain and the doctor says you know i can give you a painkiller it's harmless i can give you a painkiller i can treat the pain directly you know other things being equal i'm

2:09:20 like yeah i i i the worst i just want the pain to be gone right and again that seems to me this sort of um self-effacing if that's the right word self-eliminating character i think it's one of the phrases i'm using that paper that cellphone this self-eliminating aspect of um felt unpleasant seems to me an interesting and highly specific feature of these states so is is the danger here i mean this seems like a a byproduct of evolution so the ai systems we build may not have this component but may still fill unpleasant states but we may not have built them in such a way to be motivated

2:09:52 by them is that yeah i mean it's a really it's a really interesting question um i mean it does seem to be certain aspects of felt unpleasantness are constitutive in a way that i've struggled to make sense of what it would mean so a state that felt bad but didn't have any motivational significance whatsoever for me um i mean obviously you know there are cases here that you might want to um talk about like um uh patient pain syndolia is you know one case here but i mean i think this is obviously patients who are um don't seem to care about their pains right right you sometimes get induced

2:10:25 sort of something similar to pain assembly or on morphine where patients are aware of their pains they don't care about them in the sense that i'm talking about felt unpleasantness those states don't feel bad right um so the i think it's basic it's something like a conceptual confusion they still can rate them as like how unpleasant they are they can still like tell you what kind of pain is burning poking i mean so there's a lot of the things we associate with unpleasantness there right i agree but the sort of morally significant component the badness is uh the badness in the sense that the structure since i'm thinking

2:10:58 about it is absent in those cases but so you don't think it would be possible to just have pain but no desire to alleviate it i mean um so a state that was painful don't think that that obviously does happen with pain but a state that was painful and unpleasant yeah it didn't cast any motivational shadow whatsoever i struggled to to make sense of so i mean imagine if you had an agent who you know was stabbing themselves in the arm poking themselves at the shop sticking you say and they say why are

2:11:29 you doing that they said oh because it's unpleasant it hurts you say okay there's something there's something more going on here right there's going to be some other motivation because that alone doesn't make sense that that is not the kind of reason that can ground that sort of behavior so i mean another way but the problem here seems like these are facts about human psychology so why are they facts about psychology though like of you know yeah i'm sorry so i mean i i take seriously this idea that uh it's tricky to figure out which features that we

2:12:01 so i think there are some features as i said that provide good criteria for assessing which kinds of beings are capable of having which kind of biological beings that have never shared evolutionary history with us which kind which which of those kinds of creatures can undergo unpleasant states um now could you have some other morally equivalent states there's some other morally significant states i mean i would resist calling belton wasn't this in the same way but that would occur in an ai that um would be morally significant but wouldn't have that kind of specific motivation or behavioral features of our

2:12:33 states yeah i think that's a really interesting question um and i mean as i've said i'm sort of um i think there are other ways of thinking about morally valuable states in terms of interests that go beyond just unpleasantness is the only thing that matters right yeah that might provide one hook into trying to tackle those things um but i mean i i guess i would probably kind of bite the bullet there and say look to the extent that these ais uh states um had a behavioral and psychological profiles very different from the ones we associate with

2:13:05 uh that are found in um felt states of fell to blessings pains nausea fear etc in humans and other animals it's wrong to describe them as unpleasant to the ai maybe there are other reasons why we should say that it's wrong to inflict these states on an ai maybe you know maybe we have maybe ai has interests in not having its databases disturbed in certain ways for example um but that would they'd be grounded in a slightly different kind of notion of interests yeah okay well i gotta let you go because i see you have more important things to do

2:13:37 but henry it's been great talking with you and i we didn't even get to talk about conceptual short-term memory and all the interesting ways that afflicts all the things i care about so we'll have to have you back another time to talk about those things it's been an absolute joy to be here and uh great to catch up with you okay well thanks again uh so um i'll end this and just hang out for one second and i'll say bye off air awesome okay thanks for listening all right we're done cool how