Consciousness Live! with Hakwan Lau and Masafumi Oizumi

0:00 we're on air yeah so maybe I can say a bit about why we're here so it's basically over the past year or so it's almost a year I was invited to Japan to sort of have a debate about IIIT yeah and I think they kind of asked me because I have some skepticism but I as you can recall I mean is that all in the blog post last year in July I was just mostly back then concerned about the so-called most ideological aspects of the of the veal I didn't actually get into the details of the view so I

0:33 thought you know a lot of it is just how how do have the way we promote and how it was perceived by the public as well as by our colleagues but meanwhile you know I think this one year this past year more has evolved I've gotten a chance to actually see in Christoph and Giulio in person for many years I haven't been able to get in touch with them really and and talk to them a little bit and and there is also some e-mails back and forth and I've just found it very difficult to even understand where they stand and I think

1:08 I'm not the only one who said so it's becoming increasingly like that they are they're more used to talking to the public and trying to win over people who really don't have a stake in the in the in the field and for people within a few is kind of getting more more opaque and then and at this past month I think there's a paper from campaign that came out directly I may be particularly evaluated some of the assumptions of the theory so I think it just got to a point

1:40 that I think it will be useful for more people to understand what's going on so and I mean while I have talked to both sue me sound who's of course wanted to co-authors of I think the lead author of the I t's 3.0 the paper and I've also talked to Ryota can I who I assume is now work together and I find it like this is a good opportunity because I really frankly get a lot more out of understanding IIT from talking to you guys than reading the papers or even talking to Julia or Christoph

2:12 themselves so I thought it may be a useful thing we talked about this and maybe we can share it to public so opu something I don't hear you so well now the clue because you become your lesson huh okay yeah I changed my microphone and their mic okay so this is not good actually no good okay that's fine yeah so how do we proceed how do you say something before or should I just ask you questions yeah

2:46 maybe I'll just say something so I think that the the purpose for me is to me that the as you said in your email that the purpose of this meeting is maybe creating there may be better theory right if you are not fine with you know current form over iit we can think of you know improvement by this discussion so that's probably that you know constructive purpose or I mean for this meeting to me sir yeah I still don't agree with that yeah to me is almost like a danger

3:24 I worried that because of the way people just couldn't understand it and find it so so crazy or somehow it has been repented we have movement in the media despite the the active scientists disagreeing and they just been selling over and in the media and I prefer journals I worry that people are just going to write off the whole whole theoretical approach I don't think it's problematic as the specific wasted being being being couched right now so I I do hope that yeah this is going to be

3:57 hopefully constructive and and you know how my question will be exactly about shall we go into it yeah yeah yeah so I first I mean let's start with something maybe more more specific I mean ultimately I'll comment on how the more the future but the first question is really about the relationship between IIT and and an older project called the NCC project so that was also can I see yeah yeah there was a feeling from last

4:32 year's meeting right I thought that the NCC project as I recall when in the 90s you know Francis Crick was still there and uh and there was this very very kind of you know very food on the ground kind of approach and say like well you know there are so many concepts and and and even if you just talk about neural mechanisms of consciousness maybe too loaded because you're assuming that there is a mechanism and if you say neuro causes you're assuming that it's a cause or a thing cannot cause itself

5:06 you've got entities who there many of the metaphysical questions and I feel that the NCC project was started I was a very humble and neutral approach and said like yeah what correlates and get some data before we get to Soho hang-up about it and and and IIT obviously started out as a very different approach right we start with yeah we and axioms and then hopefully they're still related it's just that one is starting with the data nothing well theory so that that part I find it difficult and it particularly in my visions were people like Christoph and

5:39 in correspondence so those one point I'm it's very specific and hopefully we can just clear it out so is it right according to current formulations of IIT at any part of NCC must be causally responsive in the following way that if you deem a certain part of a nearer circuit of a certain power to break as part of the NCC according to Christoph is formally defined as the minimal set to exhibition yeah that there would be

6:13 sufficient for for having an experience if you manipulate any part of it does it mean that you necessarily causally change your experience your conscious experience I think so yeah because that really seems to me is different from the original NCC so I sign on to do this not to do something like that so so that was one is so basic I already feel like we are already diverging so for me the antiserum your correlate so because it's a mirror you can have redundancy you can

6:50 have sets of neurons and they kind of both correlate with the experience and you can say let's think of it as more concrete unless you have some neurons on unter cortex inside you oh my god and it will cover the experience and you can say well so you want to find a minimal set maybe you pick one like the lemon one left the frontal cortex is involve you are neurons there so then you want it would be the left and prevent oeko-tex then with the minimum that

7:24 determines your experience but then if you call so we manipulate the left let's say you lesion those neurons then the region right hemisphere can take up the action and contribute only the same experience so you might have causally resulted in no change in experience but would IIT a good idea like that I think so yeah you know MCC actually corresponds to the complex so-called

7:57 complex complex is a locally most integrated subsystem so IIT doesn't consider the whole brain corresponds to you know consciousness only the part of the brain corresponds to you know NECC you are in a consciousness and the contents and the levels of consciousness are determined by the activities and mechanisms in the complex actually so that's the theory of IIT and if you be on some part of the complex it will change there of course levels are the

8:31 contents but it depends on the you know a degree of region under you know the price of regions and so sometimes this may not change their content levels so much but sometimes it will you know is following the whole thing I said it's just yeah it just depends on yeah that's very interesting so it sounds like that so is it right that within the complex yeah one pattern correlates to one experience they said a one-to-one yeah

9:07 you have multiple giving rise to this exact same experience actually as shoes one-to-one mapping yeah I'm on to my mopping if they're you know the so so-called mix mixes are maximally it is well conceptual structure so that so what concentration so ietake conscious the contents of consciousness is determined are you know really integrated pattern of you know informational exchanges in the complex

9:41 and it it is called the mix or conservation so if you see our iid three points there paper you can see some you know conservation metaphorically metaphorical in conservation which corresponds to crawl in Korea and if you be on some parts this you know shape or shape of conservation changes but sometimes it does not change their you know shape so much and what is your

10:13 question well actually I forgot to get know so I said the mapping happen yeah the metal doings are one to one exactly so there are shaping one shape shape corresponds to one experience so this is sort of one to one month so if the shape changes the you know the contents are coral you know changes accordingly right but the part I don't I'm not so sure so even the the complex given this

10:48 is the NZ within the NCC is it the case that the same conceptual structure can be multiple multiple multiple realized so so you can have different allow you have different active notes that would price exactly the same conceptual structure is it possible it's kind of it seems like a mathematical question to me God mmm I'm not sure but as you can say I think that the Sun it evokes similar

11:26 patterns of conceptual structures even though you you know beyond some part if their concept just structure is similar I think the experience is more red Savina so right that's kind of you know redundancy will occur in IIT I think but the redundancy here seems to be quite radical is not just a similar so as in the case of you example frontal parietal Network we understand the redundancy is

12:00 like very different neurons can be involved in doing this similar stuff yeah in the demon you're yes so even though we supposing destroy some neurons I think that the concept of structure does not change much so that that can really does not see of course you know problems in the brain yeah yeah yeah but but here sir so this is this is good this is somewhat useful but the formal question is maybe even weirder instead once you say you you seem to agree to the formulation of first off cop in his

12:36 recent writings that the NCC is the minimum set of you know whatever substrate that is right vision for experience then in that sense that the the the NCC if you take away some notes of the NCC you can add some them outside of the NCC or example I gave periods mm-hmm all right but would that be allowed in within IIT yeah that's interesting question so I think that that if you reason some part of the brain I think that complex itself

13:11 will change so in the enterprise for example that the complexity here but if you region that for example the complex then I think that the price of a complex is not the same I think the by this region the comp the place of a complex also changes so may be that in the index of Ray let's say that this is the you know complex and the you region some part then I think that some and the outside part of the brain is may be involved in the

13:46 conference in the damaged brain yeah so that kind of thing will happen I think it is maybe some mechanisms of compensation right I think a lot more reasonable there attend and what I have new students so far but is it what you said just what you insure go you think that some mathematics can be done given your formulation to actually temperature yes yes yeah yeah so I think it will be useful because if it's demonstrated and I would think I understood that Christophe probably misunderstood two theory because he is insisting you know

14:20 that you you I remember you call it a big debate in Taipei and I funny a hard time I said yeah in some ways a very trivial debate where the PFC is involved in in conscious perception right I think if your answer would have been quite trivial if we just review the literature as empirically true voice we can clear so they cited examples where the lesion the PFC and said it doesn't change consciousness and those cases are mostly

14:56 mischaracterized or my son cited but but we gotta see what it is at the same time your theory according to IT actually the lack of lesion effects a lack of abolishment after PFC lesion actually should not count PFC against is Rowan in playing for in Sec this logically they're saying that we've Elysian this it does not abolish NCC it's actually incompatible with it with I I think logically yeah so that's interesting I would find that's a pretty

15:28 good divergence of opinions already I think that that one thing I think is constructivist because of the weight Christophe promotes of theory and as a proponent they had their version of the theory which sometimes get over represented as representing the entire entire entire you know approach so you guys seem to have quite different ideas there which is interesting I button that paper you know about you know whether the prefrontal cortex or our you know posterior region is you know the essential for consciousness it's not

16:02 based on theoretical prediction of IIT so that paper is not too there IIT I think that paper is only based on empirical evidence right so there is no say little prediction in that paper but I would say the interpretation of the empirical evidence I think firstly it based on miss reading the papers I think they they actually anatomically got the Anatomy wrong so if you make down I at least we think they're wrong we think

16:34 they define the profound of context wrong and they place the central sulcus in a way that is not what we see as the textbook definition but but regardless of that I'm just saying oh just just logically the way they interpret the results seems to be also against a very concept of of IIT I going to IIT if a certain part of the brain is part of the NCC and if you leisure here is this not a necessary consequence that you would you would change the experience and yeah the occupation is based on an assumption this seems to be against a very theory

17:07 so I know that it's not it's not based on the theory but I'm what I'm saying it seems to be based on something that is incompatible with the theory itself that is more incompatible what do you mean actually I had gone the Sun fly can I try to jump in here if you guys don't mind I think what horn is saying is that you know what you were just kind of admitting to was that it's possible that there could be redundant systems outside of the NCC traditionally defined by I would yeah which might do the job just as well so then if that's the case then

17:40 it seems that you can't infer from lesioning of a particular place that it's not having this redundant role it might be you know the the mix might have been equally well realized by that cortical area as well yeah I agree yeah I mean is that fun – yeah a lot super so useful so you understood IIT like the way you do they would not make that interpretation in the first place the fact that lesion doesn't abolish lesion to PFC doesn't abolish consciousness is no evidence against PFC as part of the NCS yeah yeah

18:15 doctor yeah I agree yeah I believe that fun yeah I see that's interesting yeah I think just it's a guest because that you know if people may guess that I you know if you region some thought some essential problem on the brain I think that the or any CC or complex I think that the people can guess that the it will you know really change the you know consciousness a lot but that but that guess what quite so educated but um the people specular of it yeah yeah this is

18:49 not necessarily true from ILP like a while we are in your favorite color viewpoints it's just I you know prediction yeah okay let's not me so I think that's useful now bellies at least I mean I don't know whoever's right maybe they would say your understanding of IIT is not correct courses clearly theoretical viewpoint again that it will happen data you know even though you you know region some part of the complex it will not change much consciousness yeah

19:24 but this is the guy I prefer your your your version of piety yeah yeah I mean just to interject ones quick a lot of this depends on on something that we can't do yet which is to really identify the mix or the conceptual structures that the brain is I'm Stan [ __ ] yeah so a lot of this is really beyond what can be delivered by the theory yeah yeah so it could turn out though that you know hoc one the prefrontal cortex isn't a redundant system in this way and then their thing would be verified or vindicated where it's also it could turn

19:57 out that pending on how you measure fire or whatever that that is another kind of redundant system so that's kind of an open question I don't think I would say you know Richard does a pretty good point but I think on almost like all the neuroscience questions are somewhat open very few of them right but yeah the the redundancy claim is is really some of the bread and butter textbook stuff you would you understand from from from you know cognitive neuroscience yeah and I would think is not quite good for business to just just to not take that

20:31 as a good assumption given the rest of the fuel already does unless we have very strong reason to think that PFC is not redundant it will be going against your anti industry Mir on everyone who studied PFC knows these we done and see you know you talk to people like Bob Knight who is a co-author so I might be partial but they are evidence to like if you lesion one part the other part takes over and does the similar functions etc so I'll end up right in front of a varietal have can't counterparts etc yeah in consciousness in general

21:05 disconcerning functions in general but if you but they seem to be a good starting point to think about this yeah I mean it sounds like part of your question is whether IIT as a mathematical structure is the right even has a appropriate mapping to the brain as a biological structure sounds like that's what you're trying to get out baby huh yes Richard you keep it being my question this is good I get back to all of these if you don't mind like so about the donor so that's another

21:38 question that is interesting so about the notes of IIT so which you just mentioned about whether it maps to the biological substrate so the notes are I think you're on tour what are they by Oracle comes right to the notes in IIT so the IT is a theory yeah yeah the notes can be anything so it can be you know of course neurons or other things like artificial things yeah but I think it's applied to the brain

22:14 are they meant to be anyone yeah yeah brain that it should be neurons he runs on no time you're on her firing rate and and you know tempo yeah yeah how does that work I it never is very clear to me so to everyone the neurons don't just turn on and turn off neurons are firing rate as family and also the some membrane potential yeah so is it is it meant to say a neuron below the firing pressure

22:48 is off yeah that's another interesting questions or how can we determine the state of note which is relevant to consciousness so I think that are you know always that the maximum of fire is relevant for consciousness so to find the appropriate state of no over not above null we need to search every possible you know state every possible candidate on the state of the notes like it cap if if you consider neurons it can

23:24 be you know maybe the firing rates or someone each spike or some threshold you know membrane potentials of activities it can be anything so we need to test all other possibilities then we need to charge the maximum of five over there all possible candidate state then then we determine the relevant sort of state of you know that system of the system yeah so in some ways we didn't known

24:00 physiology we already know that the local field shows are important the firing rate value yeah and so so neurons are not if you take given a single neuron it is not to say what is yeah it can be you can be firing very high and have showing relatively less you know local field potential and a certain frequency and something so so how how do we go about determining the state so it's sometimes that the one

24:32 neuron at multiple states yeah and also we can think of maybe multiple neurons activities right yeah we don't need to consider just one neuron maybe multiple neurons can you know be little ant to consciousness oh yeah that's right that's right but whose mother let me just concerning up it map one neuron to one state it's it's yeah is it it's in the case it is more likely that one neuron is represented by the graph in IIT it would have ended by multiple multiple I mean that one node

25:06 can be multiple neurons right we can think of that kind of thing like oh one can be varmints diesel who are on could it be the other way by the mighty one neuron can be multiple notes yeah one your kami yeah I of course is actually yeah yeah if you think of the for example of the miraculous huh I mean that nickel levels you can think of it in that way but now what a minute they're given the physiology it looks as a one neuron definis more than two states who almost like anything if you

25:39 have to map it one near Anyang oh notes yeah but then it kind of suddenly it gets very heat Harry I mean so yeah yeah yeah so yes so it has the kind of structure that is about neural networks but it's not neural networks anymore these notes are not neurons yes states that are not the same as neurons and and it's not clear we implicitly assume that the one node is corresponds to one new but so logically with Emily yeah in theory actually that

26:14 that we need to start all possible you know candidates all those things so it's concerned about micro micro you know discussion so so that we never know that which you know spatial scale all of which you know temporal scale is relevant for consciousness right so the IIT thinks there is optimal real special skill optimal time time skill which you know generalize most information or most integrated

26:49 information search this space how is it an are you suggesting an empirical project and searching or how are you gonna go about doing it because yeah I'm not so sure exactly yeah it tell me down to some extent empirically like you can change the scale spatial scale from neurons to for example that the some grindage or the macro scale or something or you can test field what in in shalwar other states then you compute

27:26 the information will integrate the information then you can compare there some barriers then you can find some you know appropriate scale war likely to be appropriate skill this is interesting sir so let's see you computed a scale so the different scales as you go to the ion channel scale then you get one reaction yeah yeah which is that lesioning prefrontal cortex does not change consciousness in that case and then you have a different scale which is this a

27:58 proper area level one problem area when you're actually in a together yeah if they saved a truly self stone converge according to the theory uh-huh what do you do they take the maximal what do you mean so you analyze things by one by one simulation you submit them at the ion channel and then you have their you predict it okay in that case lesioning prefrontal cortex will not matter so you rather never it doesn't change your v

28:30 very much you drowned I see and then I think changes you find then what do you do do you prefer that or you prefer the one that just you like better somebody else I mean we cannot test this theory empirically right like we can't you know make for example all of the neurons are all of you know microscale activities on the news let's say you can and let's say I've given I give you all of those things and the hypothetically I can compute

29:04 everything they're not fantasy let's see here everything and record everything I still on there if you then get usage scenario yeah you can arbitrate them based on their pure core results ultimately make on everything and complete everything then the theory theory we can find the sort of maximum of five an action movie from information then we've had some you know appropriate scale spatial scale on the temporal

29:38 scale yeah we can you know theoretically compute the level of the contents of consciousness in a given system then we can do the prediction oh you know how the region changes there are level of context of consciousness in the given system so I think that the it will give a unique prediction based on but but then then these different these different IIT at different levels would give different with a country with data

30:11 basically he's actually yeah yeah exactly yeah well yeah that's the exclusion postulate right so that yeah yeah so if the lower level system has a lower maximum Phi then it's the right scale is the one where the Phi is maximal exactly yeah uh I'm not sure hanging hemorrhage it I'm confused on your question here it's like what I assume is how I was saying that it's not based on principle that you choose stuff it's based on how it fits with empirical data that if Vista the two things seems to be different is that right not three yeah what I'm talking

30:46 about is just purely theoretical stuff it is you know this is all related to empirical I see I see so you have different you if you draw out the IT Network to compute the Fayette cetera a different level yeah Avante yeah yeah we should pick only one the the iit postulates that is actually exclusion you know principle says that the only the maximum matters

31:21 for consciousness other simultaneously but of course it is debatable maybe they are going to be multiple scared of the consciousness yeah how come Mitchell yeah are to the axiom of exclusion that if you if it turns out that you're you're to the level at which you can maximally exclude the others that that turns out to to give you empirical evidence that doesn't seem by right would you go would you be willing back to the axiom and and change that or what

31:54 we add up that's to question to me yeah yeah exactly so let's see let's say you turns out that the level that give you the maximum 5 was was at the ion channel level aha and then empirically we know that that doesn't seem to work out so well you could take advantage of it best food ion channel level notes it doesn't change your politics such ad as effectively as some are at some course at grade levels yeah what would you do I mean would you well you know we need to change something of course yeah it is you know

32:28 compatible with the experiment evidence we need to change the theory would you be willing to even go so far to change the axiom itself to abandon this yeah I think another by the way sauce the name of actions seem to be in appropriate right because the people are sort of not comfortable with this you know use of terminology and pounds axioms actions sound like mathematical actions or actions in physics and I think it is totally different reality I think that

33:03 we should just correct as a property of properties of the consciousness yes so or essential properties of consciousness and this actually is important because that this determines scope of the salary and also the determines the definition of consciousness based on this server so the IITs sort of first declares that this is consciousness like information on the integration and excursion these other properties

33:36 IIT considers as consciousness that's sort of determines you know just what kind of program is it tries to solve and this is I think even more important parts for creating any kind of theory of consciousness otherwise that you know the definition of consciousness is ambiguous right I mean that many people can think consciousness as a different things but first we need to declare that you know what consciousness is hang on I think I think that I think that's why

34:09 most people are particularly uncomfortable with the axiom of exclusion because that isn't much about what consciousness is exactly the property of the wexel so the exclusion is most I mean that the controversial nice yummy ID I think information integration I think these two parts are sort of acceptable right to many people to some people yeah I think you can not not all people that to

34:42 some people but the exclusion is a really unique I mean unique unique sort of property I mean idea saying so and also it is the most interesting part for me because of the we can test we can sort of test empirically test this you know exclusion principle yeah in AIT for example that by using our spirit brain case it's interesting you mentioned that you know

35:13 these axioms are somewhat different from mathematical axiom because yeah exactly who you talk to in mathematics some people in mathematics would kind of say axioms are no longer referring to the same kind of you know Euclidean dasor we use active items I kind of have policies to them but of course mathematical hypotheses are very deep yeah yeah very complicated industry and then proving something is a very different from what we take is proven science and yeah but would you take the

35:47 same like your Exum's are basically hypothesis then why don't we just call them hypothesis yeah I think that we can call it hypothesis as well yeah so ultimately exactly I think it sounds like that you are essentially an empiricist at heart to the extend that if you have an axiom slash hypothesis that doesn't turn out to work if you if you if you try to do this axiom of exclusion and run the by certain you know system by the time you can compute and measure them it turns out that in it is sense that a glass of wine is conscious and you are not then then you are more likely to reject the axiom than

36:21 to yeah exactly yeah yeah we can call it just a hypothesis yeah there used to be called postulates right I mean yeah Australia's other good yeah yeah yeah yeah I mean I think this is a really important point because some people think that well you're you're more married to the axioms than you are to empirical data and so it's an important corrective I think that what you're saying should should be told to more people because what you're saying is that the the axioms are responsive to empirical data and would be refined in

36:55 light of that which which makes it more of a scientifically sounding endeavor whereas other people would say look axioms that sounds like you're not even doing science empirical science you're doing something else I think this is an important kind of clarification yeah and in a way it's also important to say like inside so that you and Julia are very different then I think probably the way axiom came may have come from him I don't know but but he certainly uh he was your advisor at that point and he he speaks European languages so so I I figure that the idea my company but he

37:30 he actually I think he stood by them as real axiom so in the Indus Aaronson critique when he was at some point when when Julia was forced to the point that you have to say an inactive set of logic gates a conscious act Julia would say yeah because the theories that this is and those are from axioms and so it has to be true at least the theory is true then the rest is true rather than the other way round he's he seems like he's more really from an axiomatic approach you might be more empirical than than not really actually that the induct prospect

38:06 in that point I think that I'm same as Giulio yeah because at the we first need to assume that the series lights and then that the logic gates is conscious that's our series prediction if it is not true of course that we need to change the theory that's how the things well how are you how are you given ever gonna know in actual logic gates are conscious no no yeah we count you know of course that if the theory end up saying that you are not conscious you would then right in that case that that of course

38:40 we need to change the theory so there are cases are first we need to test the celery in humans or in animals which we think that they are conscious right then we accelerate the results of the theory to artificial things that's the order of how he knows that's always tested yeah which brings me to one point I mean so to test humans of course human brain is less complicated and right now

39:13 you can do so at the last esse in Poland and embarrass thought ups of these issues make that note of the test of theory you have to approximate and I know the approximations don't really converge so different schools are the best it give point boy in a different yeah yeah yeah that seems to me is a similar problem to the level problem but this time it you don't have the exclusion criteria to to see like a priori which is better so when two different approximations don't converge what do you do yeah that that's

39:47 actually a very good point so it's sort of related to the action things I think that the in physics that the from axioms or postulates we derived sort of unique prediction we can derive sort of unique communities but in IIT only from actions we can derive unique you know quantity which is integrated information so there are many definitions of there are many possible definitions of integration where

40:20 information so one is of course proposed in original IIT and the also the other conduct proposed by me or other researchers so there are many possible you know candidates which may qualify the concept of information why integration so so it is it cannot be uniquely determined that sir I think the deep problem in IIT consciousness there should be some unique karate right

40:53 because of our consciousness seems to be even it um you know unique properly so in the sense that the some unique mathematical quantity should correspond to this in our consciousness but we do not have any principle to uniquely derive that cannot be right thank you feel some of them so in in other barons talk yeah he started to distinguish between two kinds of IIT like five measures and they called one of yeah so

41:27 what is the Imperial so are you are you also kind of on board or sympathetic to that distinction that that seems like yeah yeah so you are as well yeah there are two to you know major difference in five candidates so far proposed in literature so why is the original one original man actually tries to qualify sort of actual causation actual causation in the system not a you know an empirical observational causation really physical actual

42:02 causation and they you know the IIT tries to qualify this you know physical causation as I you know information or in information but uh in the you know empirical file actually that far we can do is just you know observation observing data then we from Eva there are seeming causation which may not be necessarily actually no physical causation but even if you do participation if we can do perturbation

42:37 we may be able to qualify for backshell position which is sort of similar to the concept of ECI you know uh debate round as well complexities like I don't need a motor vehicle PCI still very highly approximated highly appreciated yeah you wouldn't say PCI is a fundamental measure either not not fundamental of course yeah I do not necessarily need a need to fight to sort of integrate the information itself yeah PC is a sort of a you know similar approach of – IITs you know causation codification yeah but

43:14 I I mean actually maybe I'm more of a spitter what is it I don't know he froze Pakwan lumper I froze oh yeah you get out am I here yeah yeah you're here yeah yeah yeah I was saying that I may be more versed bitter than a lump er so that people would say like you know wise and I and I worried that here the distinction between the the real fundamental kind and the PCI type you see they are similar but you know

43:49 everything is somewhat similar you can also say IIT is similar to global workspace yeah I worried you know theory says being inclusive and all that but but here be given the you know some many many aspects of IIT seems to be pragmatic is it more useful to say like to to to to to try to flesh out the different versions of it and and see which which of them might survive some of the criticisms and some of them may not yeah yeah we can do it in this way sir uh right it's that there are many possible

44:23 you know candidates in debris deflation but some some may not be you know counterintuitive or some maybe you know intuitive I mean be correct so from this kind of you know test or from this you know argument we select the you know so appropriate measures of integrate information so that's a current stage in IIT durometer mathematical viewpoint but then so I'm still use about this so so if you start

45:00 with the axioms and thank you mo almost like postulate of something that is very likely to be true and you ultimately try to see how far the theory can go and then is it so so when you have different versions when it folks when you have different versions of the theoretical approximation though but PCI's tells you one thing I know to be judging which approach is better by data or I see that the we can mathematically discuss the which measure is better fast and then we

45:35 am very very you know evaluate which measure is useful I think yeah but what if they go what did they go against each other so what if you are Peoria you have won one measure or one level of one approximation that you think is mathematically more elegant or better but if it turns out that measure you know we did wrong we have got what the people gonna gonna come back in the coma ah and what do you do they of course that that we need we need to you know

46:08 discard this you know mathematically good measure of course yeah we need to reconsider the theory to serve right yeah I'm not so sure why why that is the way that you you prefer the mathematical properties so much I understand you have a physics background but how about in an email I think sometimes miss miss applying this physics approach is a bit dangerous so physic come from a very empirical background right so yeah they quite low like status not because they

46:41 are mathematically elegant because they yeah you actually work in your work but yeah of course not yeah that's true but the first that the full construction I mean after you know construction stage of the salary I think we need to forget about empirical stuff to some extent of course we cannot forget everything but I think that the if you consider all of the you know empirical data all of empirical evidence that is just a you know sort of over feeds of the theory so you could have

47:14 just over feet it's every we we want to have the more general salary right otherwise that the Newton's theory cannot predict for exams and motion uppers and motion planets so it is really general because that it is January is constructed so fast that I think that they start from very basic you know minimal set of you know for example the equations or principal actions right then we can deduce all of

47:49 sort of general things like from motions apples to motions of planets that's how the theory is should be constructed this is security really I wish I don't know which mission probably we'll know more about philosophy of science but my limited experience in trying to avoid overfitting it's nevertheless data I think consider a month later not does not help you do not over fit in fact if you try to look you would try to train

48:23 the model on one part of the data and see whether generalized your other parts of data yeah doctor I think that's sick on the stage I think so first you need to start from the sort of minimal set of principles then you of course need to test empirically this salary by using all available data if you find something some errors then of course that you'll get in trouble and you need to be considered a sell it yeah yeah I was gonna say sounds like what you're saying is that the state of

48:58 consciousness science is not like Newtonian physics but rather more like super string theory or something where we're talking about superpartners and we're so that's all coming from an assumption about naturalness and something that's very theoretical and the theory is then empirically tested but first you have to have something that's incredibly general and that's what you guys are hanging for yeah I think so well just that stage now is that before Newton right basically before you know it's early so we have no theory at all basically so so we need to

49:33 discuss that how we should create a savory right how we should create as a mathematical theory and the IIT somehow tries to do it of course we can discuss this and it is really debatable and I think it is constructive you know but it almost worries me that you guys have tried to create something like a super-strength theory yeah oh yeah yeah yeah you know Newton Newton was not nearly as ambitious Newton enters you know the relationship between forces and

50:08 accelerations and and that's all empirically grounded so it is a nice they started with a very simple almost undeniable phenomenon yeah it's not obvious at you know at that point that's ducting you know you know video actually that the for example that there it is really deep if you look at the Newton's you know axioms right people cannot be in all of the things at that time of course of the now we can see all of the

50:41 things in the textbook under we can see these things as or empirically you know proved and also many people are except except all of them dinah'll but at that point I think that the it is not that obvious right so obvious at all I mean that you can go back ever hollien debate we cannot imagine that this simplification predict all of the motions from operas to planets I mean that the point was that

51:18 so the Newton case Newton came up with these equations by going through all the data collected by Galileo he didn't sit and try to derive them ah Prairie he tried to fit it to the data by being better not for carefully to tailor the equations till they sort of masked what the data says as opposed to starting from first principles and deriving something out to be later you know I think that's what the contrast here is between where do you start from first principles or from a lot of a ton of data yeah I would say Newton is a through-and-through empiricists I mean

51:50 of course he invented calculus yeah women mean you know demonically clever you know clever thing that you can actually do you know differentiate both sides of the equation and then get from the first law to the second law to the third law and all this stuff yeah those are there's a match so first of all those are matters are not approximations is probably difference and each law turns out to be empirically true not turns out to be true just we can say this is you know useful for explaining

52:26 data so because of the better theory you know came out right after Newton's for example I'm Stein salary or quantum mechanics so these things are necessary for explaining function Niccolo big micro scale behavior of you know atoms and molecules oh also that if the you know sphere is really high that we cannot use that Newton's theory and the weaker abandoned the concept of you know absolute you know time or absolute space right that's how the physics you know

53:01 develops so I think I think I remember but that but that moment Newtonian physics collapse it was it was seen as a paradigm shift and yes though it was such a and the whole and the whole time that even when Newton established himself the crowning moment and all of those moments were not created by anything but empirical challenges so Newton acquired the law like stead status book forth laws it caused a lot of racism and also the getting that maybe not say the

53:35 downfall but the paradigm shift extending it to two quantum mechanics and relativity but also driven by empirical it's not even by empirical evidence for example in the Einstein's case that people are try to make other sort of just a modification of Newton's laws right then actually they fail basically so they need n stein zero sort of paradigm see like a science or abundance are you know concepts over

54:10 sort of obvious and it's time on the space and then he sort of got a fast you know that there assumes are some principles like for example that time the speed of light is constant that's the principle einstein's causes when day he creates they're in a theory but other people never think of thinking that way right well initially because if you use the Newton's row that the speed of light can't be for example at the constant I

54:45 mean I thought it was part of max comes out of Maxwell's equations at the speed of light is like that and so I don't think he didn't I'm Stein didn't come up with that he just kind of took it seriously took heed City also to use actually yeah but but it was so I mean so the discussion here is very interesting because it sounds like on the one hand we have theories that start from data and only the data and try to build from there and other theories that start from some some more mathematically grounded things and try to build back to the empirical data and you've actually

55:18 yeah and so I think hoc ones question is why should we / why should we prefer one versus the other given that we're trying to explain the brain we have a ton of data we should be looking at that I thought that was I kind at one point they are even spawned we never you know the history is all over the empirical evidence of course what I'm saying that the fast let's try to build a more general theory so from some basic principle then we need to test

55:51 this general theory with the order of empirical evidence on empirical data that's how the theory should be created oh that's a hot house the theory has been created in physics right that's what that's what I'm saying I'm not sure I agree entirely about I'm hoping I don't you know this is this is very useful in in in Thomas Kuhns writing and say that well we look back at it Oracle chapters from high school to your college level to your graduate level

56:24 physics as if he's illegally progression but of course PDF on the nails over over those things and and the history of his physics is actually a lot more empirical than we might have thought to begin there so that's one thing I would think is probably I read but but on top of that that that's a big piece of history that we can you know go back and ask more experts I'm yeah yeah of course I'm gonna thanks I just I'm sure but but i but i but I won't worry is also this in

57:00 this analogy here in physics the prediction the empirical predictions are always gonna be quite clear yeah the electron to a guy okay exactly yeah and in our fuel they were ultimately cases like in active logic gates do not you Leo think that they are conscious I think that's just ridiculous but I also know where to disagree with him so our empirical basis for falsifying some claims are if we endow a little bit even more than what we already did in physics

57:31 I worry that this is just not good for business maybe given that we are empirically on shaky a grandstand physicists are or were we need to have been more more attention to the empirical side that physicists did yeah yeah that's totally – and by the so the for testing the theory that the fun we can do it's just this was an approximative approximatively testing the theory right the physics there are power of physics is that the sort of accuracy of prediction they can

58:04 predict DV accurately predict the point of the motion of planets or the you know how the three micro scale particles you know move are these really really accurate prediction but in the field of consciousness I see that the the prediction cannot be you know accurate anyway right so the I think that when we discussed the theory of you know consciousness that the [ __ ] I think we need discuss it from the viewpoint of

58:37 usefulness I think or in physics also that we should discuss the theory whether this is useful or not because of I when we think about the physics that that we think this is sort of fundamental theory well so the law of you know nature law of the world but this is actually only the approximation I think because that the the the history proved as history proves that we can think that the better and the various theories and physics works so the but

59:14 the whether the theory is good or not should be determined at the operator the servery is useful or not right if the theory is you know predicts in you know to some extent the predictions that we can say this is useful and we can use this theory and also that I mean consciousness that is the theory predicts something useful I think we can you know accept this or we can use a sort of white barking hypothesis all right I entirely agree with you actually

59:48 yeah that's also why we started doing more and more clinical work in the lab on night unconscious exposure therapy yeah yeah I don't know what you think like so the strongest use of IIT so far has been got a PCI relates it to put the patient COSO index is right PCI yet yeah one thing is that you know determining the levels of consciousness in humans right in vegetative patients or in coma

1:00:22 patients be honest eyes you know page answer yeah you know useful application yeah to me suing the strongest suit of a somewhat heavily promoted aspects of IT but I always worried that are we sure that if we do just machine learning we would not do that yeah I think that we can do it in that way yeah but if you do if you do machine learning and it turns out to work better there must machine learning sometimes can't have you know sort of fundamental understanding of

1:00:57 camp you know Mia hey machine learning can be applied to anything like for example that the goal for example and we can create the you know better than humans you know that the AI in go watch air so any kind of game but it does not mean that we understand oh we understand chess right yeah only a I can do many things but we do not have any knowledge about you know go wow yeah I have to applies to that what is it well if the

1:01:31 point about selling the the criterion of just prior theory is its usefulness then I'm not sure understanding is that important if ultimately discuss yeah understanding is that facet that usefulness is the important aspect but also repair we want to understand consciousness more fundamentally right that's it's two aspects right yeah yeah in this sense there is a second aspect which is actually it is I think it's not a fair statement to say the machine and it doesn't give us understanding because

1:02:04 once you find a machine learning model to do something you can then go and so-called dissect model you can do what is proposed you can actually test the model you can you can run Studies on it my colleague at UCLA and the beach they have you can you can turn a deep Learning Network back to a decision tree so there are different ways you can then understand from there what I want worry is if you start with machine learning and start to understand how to predict where the patients would come back from a coma you're going to have understanding they're very

1:02:35 different than IIT I mean that I answer that I'm again not export of machine learning as well but my question here is that can we creates the physics of theories from purely machine learning methods right can we you know create Einstein's law why Newton's laws work on the mechanics based on only machine learning of MP dot data that's for my question here you say I think you answers it well yeah that actually leave

1:03:08 us leave me alone think about can can I ask a question I know I know I'm not an interloper here but I'm wondering um so I wanted to ask about the axons and about the usefulness so you know take a coma patient or and trying to distinguish locked in so called syndrome from traditional coma so suppose you are able to look at a person and you had a way to estimator useful measure of Phi and you say okay so there's a high level of integration in this coma patient so

1:03:45 therefore they're conscious that's the kind of inference so then you know and you didn't want to do the same kind of thing like in split brain cases they were all right side of the brain has high file oh it's conscious so much so my question is is isn't isn't there a problem here with that so the theory that the axioms of the theory sort of seem to be geared explicitly towards a certain kind of theory of consciousness that we call first-order theories what in short say that you don't need to be aware of the of the internal organs of

1:04:17 your mind in order to be conscious so so isn't that question still left open though by these kinds of empirical results so there would still be a question so even if you found this HiFi in the right part of the brain there still is a question well is it conscious because when the subject is not aware of it they can't report it so it seems like there's still this other question about how we're justified in inferring the level of consciousness there if we haven't already addressed this question about the dispute between whether it's some kind of inner

1:04:49 awarenesses required for consciousness which the axioms are silent on or don't say anything about so I'm just wondering what you might say about that I think this question is related to Grover Locke's theory so I let the sort of vast paper and I think that the in global workspace theory that they finish on consciousness is like that consciousness is the portable thing so accessible yeah right and IIT does not consider these

1:05:25 aspects of consciousness at all actually right if you look at the actions idea or just property of some consciousness IIT it does not say anything about the portability right that's really huge difference billing as in IIT and our euro roba workspace every so that if you think that the consciousness as the portable some you know things going on in the brain that the idea does not account for anything any kind of this at

1:05:59 all right but so then so my I mean I wasn't thinking about global workspace for more like higher order theories but the point is in either case the question that I was wondering about was how useful is the theory then they can't distinguish if it's obvious us as ah I guess baby the question against the other side in a sense by encoding or the axioms not that theory so then you go AHA look there's this hi fight over there there's consciousness and then the other side goes wait a minute there's

1:06:31 still a question about whether there's consciousness there so how useful is the theory then it and all these other issues are decided okay that's a difficult question so if you if we did not do a portable we cannot say whether this is useful or not right there and your question yeah yeah that's so that I think that there what we can do right now it's just testing the theory with our sort of come on come on over come on sort of consensus in

1:07:07 ambiguities like we think that for example that the in coma patients regular level of consciousness she read oh that's a sort of our assumption here and so first what we can do is just testing the celery with some easy cases or some commonly acceptable sort of cases right yeah I am not so sure that those cases are damn in there yeah interesting because it's exactly like even your one causing there in the gray

1:07:41 so challenging ones easier ones almost like the one that Joe LeDoux will talk about if you give people certain kind of anti-anxiety drugs and then people the hand stopped sweating they stop having butterflies in the stomach they're hard stuff but if you asked him I used to anxious is like yeah do you like going to say no thank you is used objectively you dread and and to me the case of where the physiology and the subjective reports really diverged and the usefulness of subjective recorded access become very clear if you

1:08:16 really want to be useful it's not good a sailor well actually your fire is now very low you're you're basically in a coma you're you're not suffering you're not in pain this is the best anesthetic you can give and and then the patient starts screaming and saying I'm in a lot of pain then your fear is not very useful of course of course it is not useful at all yeah important useful to him does effective report seems to be really important aspect I see yeah that's that's that point I think yeah but arguing that the as Isis that the IIT the

1:08:50 is not necessarily related to subjective reports because of from the construction of a theory right it does not take account of you know the probability why don't we I mean we are yes you said that you seem to be much more open-minded then then that some of the colleagues and then there were I thought of you so why don't we create a version of IIT that is taken to account of complexity about your network is actually to try to say how which with reportability might come out from that theory yeah I think that's why index point and I actually

1:09:26 need to think about this because at the when I did the global works for theory but again that I noticed that the definition of consciousness is actually portability in global box facility right that's what they know actually as far as I understand yeah but the what is vacuuming global box every its massive massive mass mathematics I think right now I think that the they a group of special needs son you know mathematics – you know

1:10:01 clarify what for example that the globe are sharing more information means or a fact even if you means are you speaking of mathematics as in computations don't palace that medics so you really want as in as in mathematics as in computations or not mathematics so you're – not mathematics actually the mathematics is usually for I think the clarifying what something means would you take a computer simulation as

1:10:35 mathematics no no I can't be the summation Massey Massey III yeah I consider it it does mathematics but it's the work space is completely mathematical because now you get yeah I think that the more my equation equation exactly yes we analytical solutions not just yeah you know iterative simulations yeah not the various missions yeah we don't that Gruber works best so GWT GWT also considers information as some important

1:11:12 aspects of consciousness right because that information needs it to be you know grow body shield – you know generate sort of integrated consciousness but they do not you know clarify what information is well what the information means right know in the mathematical sense not the mathematical sense I think he was yeah it was a clever I think I think most computer scientists wouldn't be happy with their with their

1:11:45 formulation as an algorithm of an architecture or something like that it's just that you are saying yeah yeah you just said it's not you cannot write it down as in equations but a model a lot of computer science cannot be written down as equations either they can be only be written down every firm yeah but for sailing you know being tested in and empirically I think that we need some equations because that edicts on Korea which part of the brain is GWT from the theory right

1:12:18 no no no just empirically you can sort of predict maybe this part could be global waxes but what we need to have what we need to want to have already mentally is that just given a bunch of you know newer data there are activities of neurons let's say we have all over the recorded new neurons in the brain then which part is you know Gina global wax this that's that the prediction we need from the celery I I mean I'm not

1:12:52 sure I understand it right now maybe those predictions are not made precise because the data is also not so precise but I think I'm interested this philosophical point like if you have a description of an amber you have an implementation you have simulation yeah that is precise enough that that is basically beautiful yeah yeah of course that the simulation is okay but simulation that the how you can determine that the praise of global walk space race of global walks face Reiger whether the boundary of global box face

1:13:27 like it's kind of a similar situation to IIT like IIT you know tries to identify the phrase of consciousness as I in on the phrase of you know complex and also I think that the JWT identifies the phrase of consciousness as I you know the trace of drama box space the very global boxing base from the theory right yeah I find it yeah so so I would have thought the way that Google workspace works is exactly in concordance with how

1:14:01 the whole industry of cognitive neuroscience works is that you have you have which is actually based on you know God of course cognitive neuroscience is just based on cognitive science we can think you can take David Marr as an [Music] archetype you know model for that and and and you would you would then first of all describe the goal of the algorithm describe the algorithm and describe how it's implemented and then try to spell out how its implemented as as much as you can it's not always equations sometimes it's similar like an interactive algorithms that the

1:14:37 analytical solutions that may not be clear but once you have those you can then start to find out the boundaries so in fact the rest of entire field of cognitive neuroscience you think like working memory so where's the boundary of working memory and yeah yeah we kind of do you whatever from and you try to see which of these neurons seem to be implementing those algorithms and then you would say okay so the hippocampal neurons yeah it's probably not the bound not part of working memory because it doesn't seem to be doing the same temptations and that's that to me that seems to be doing quite well

1:15:11 whereas variety the problem seems to be quite a lot of orders of magnitudes worse because of those are month blows and and approximations because of the way but but by their so that the if we determine the tongue you know definition of information then we can solve uniquely he's done in the praise of consciousness according to you rely on approximations and you also don't know what you should approximate them at so there are there are mathematical protons before you pass

1:15:46 them it seems to me that the gases are very wild whereas yeah the case of bulwark space even though you know you're not exactly sure which is exactly the boundary but it's the principles are clear well you take this neuron you work out the algorithm and then you try to see whether this neuron is part of the likely implementation of the algorithm and and that seems to be you know it's not a precise exercise but this is within the realm of cognitive neuroscience that that this is now have identify anything useful so far for the brain whereas did the project of of IIT

1:16:19 seems to be quite a bit different different yeah everything is right it says it you're not identifying a function so you don't know the rest of the knowledge of neuroscience is completely opaque to you that's yeah yeah I think that I I agree with that point a group of expiry is constructed from the in according to neuroscience evidence of cognitive neuroscience and but also that I think that the Globox theory takes account of some notion of

1:16:57 for example global sharing of the information and a friend the information is shared globally in the brain that the consciousness emerges but otherwise that it is unconscious that that's a house theory predicts but I still don't understand but for the global sharing means exactly so mathematically if they can define the mathematical information

1:17:30 could the notion of global sharing we can test our ha this is when that consciousness emerges well this is where the you know activism brain is unconscious yeah as you know we can sort of empirically test a prediction of salary if we have some you know mathematics in that science but you know as you know I'm no fan of blue board space theory I have been equity you I'm no farmer i I've been a critic of it for most of my career actually I so recently

1:18:05 I spend and I co-authored our very first paper just this year and in which we draw the line there is C 1 and C 2 and you know okay yeah but just to but but just to defend them here for the sake of it I do think that what you just said about you need mathematics to understand the information sharing I'm not so sure about that because in a sense we understand how information is being

1:18:37 shared on the Internet we understand that what it means when you say some soft parts of the network some modules can access a piece of information such as my critical information yeah I don't know whether you can write down equations to describe the following phenomenon which is I have my clinical information and I'm trying to buy something from Amazon and if someone hacked me and broadcasted information that everyone can access information everyone can use my credit card that seemed to me a very simple phenomenon that if you don't if you're an ocean cannot understand it it seems

1:19:10 to be so much the worse for your notion of information yeah and and Computer Sciences have no problem understanding that it doesn't mean that they can write down the equation to describe that to your satisfaction though I I'm not sure that I am assigning fluid space me the work space okay parts of the brain yeah and some parts of the brain can access a piece of information and make use of it just like you can use my credit card number if I give my credit card and then some of the parts can understand it by

1:19:43 global sharing it means that every every support can access this information I can use it access what does that mean like other parts can actually take information to buy things yeah by things okay the transsexual execute resection yeah glory maybe one definition of informational global sharing but we can sort of mathematically formulate this notion and then we theory you know compare it with for example the notion of information in IIT or the notion of integrate the

1:20:17 information RIT I mean that just inverse we can't completely compare the prediction theory but if we have some mathematics we can sort of you know compare them like IIT and the global box specially in you know same ground I'm not sure I mean the that turrican science fiction may be funny like the way that we are not talking you know in different countries over the Internet is built on this understanding of information yeah talk about information

1:20:52 being shared globally or locally without invoking any notions like IIT yeah I mean I I'm sure that there's some mathematics behind information as in you know Shannon's information theory yeah yeah take a notion but by and large when Computer Sciences or maybe even though they comply scientists they say software engineers talk about information being shared from here to Japan they do not invoke notions like IIT they just talk about like whether you can see my face

1:21:27 that is a piece of information whether it's being shared is shared why why is the lack of empirical formulation hindering that progress no no just uh I need a sort of codification yeah right then that for example that the difficult but faster faster to to construct a theory that the first we start from the sort of properties of consciousness

1:22:00 that's uh IIT right IIT starts from the you know crowbars consciousness and the end of the what I IIT tries to solve is that first the balance of consciousness yeah which is that the given a you know system that the to what extent you know to what extent that the sale and the system is conscious and our IIT you know actually relates the integrity information to levels of consciousness

1:22:33 and I think that the global box space theory actually does not take account of you know the devil's of the consciousness I think they do they would say that your your global broadcast ability is global broadcast ability like obvious is that if the global sharing is sort of 0r1 phenomenal like it would be exactly like how we understand the internet so we we

1:23:10 speak I mean I I'm not an executive software engineer I do work in a blockchain company as I III don't think those guys ever tried to write down exactly in in mathematical first principle how how information is being shared in on the internet they would say internet right now is the backbone of so and so you know bandwidth and you can share information so right now isn't that that to putting another way how much information to share should be quantified I think it's qualified if I

1:23:43 fight by some very crude you know measures of bandwidth but not by first principle of mathematics this is how Internet is going if you ask me like what is the what is the information sharing ain't that it that of course we can quantify information right like uh like a bit you know in bits like how much bits are Shia or something but the difficult question here is that the far east relevant information for consciousness but but why don't we check

1:24:16 the information that we talk about every day and start with an analogy there I think it's it's the TV it's a I think that it's I think most important things in voice of IIT or in the you know consciousness researches but it's information there about consciousness I think localizable David Chalmers also argued about some you know information probably is relevant for consciousness right is it that yeah but what what kind of information is

1:24:51 relevant for consciousness uh I mean that the channel information does not you know take a condom for example that some quality or quality of information or something okay yeah yeah I don't I I think I'm truly very aware of the fact that I'm getting out of my comfort zone just going into the computer science engineering of it and clearly this is something much closer to your heart then

1:25:23 anything else about about the nature of how the quantify information but I'd my may be much the point I'm pressing is the everyday layperson who understand the internet which is pretty much everyone I know do not have any of those mathematical notions and and if this is stop them I mean I mean I mean just everyone you pick on the street all of your friends who you grew up with think everybody has no serious mathematical formulation of how information is most of my parents do not know Shannon's information theory but the following statement which is information

1:25:59 sharing in Japan is more global than let's say in a village in China in the sense is more global yeah yeah I think everybody shared that notion everyone knows what it means if you talk to a software engineer they would say of course off because you have internet everybody has Wi-Fi is being shared more globally and if you go to a village in China is not being shared so everybody knows what it means and I think the board space exactly is on that notion which does not realize I think I'm doubting that the everybody knows the

1:26:34 noise the information I think that the it seems like everybody knows that know some information but we could have you know different you know for example that the codification over information or one ways of course is Shannon's you know codification information but there could be others yeah but now why why do we so hung up about it if everybody can agree that this is already a useful notion so when you when you say that in Japan the

1:27:11 question is that yeah let's say that you have some empirical data and then you need to determine whether this is conscious brain activity so this is unconsciously activities based on chromevox space oh wait i would just go to any social network analyst and just look at any kind of them any notions but i think when it when it comes to a case that is very clear I would I would take any that is coming that used you can do data mining and social networks there

1:27:44 are a number of empirical indexes that are out there for looking at how information is being you know transferred in a cascade in the social net it's wrong index right they've an iterated index or you know determine the globally global sharing no information if this comes from you know social network scientist it is okay just anything Korea even is shown mathematically but they are pretty clear why don't we consider that's okay that's

1:28:18 okay so if it is clear I am okay I'm totally happy with that I say doctor I mean in them in the paper in the literature of growver box faithfully there is basically listing and I'm not sure that is fair because already yeah I'm really not sure it's fair I mean I said I'm not that good space I used to cheer when people criticize the theory but they're not being clear is not their inability to be clear they just don't find it's

1:28:49 important to to be clear because if you ask my mum if you ask my parent my they already know what it means like when the information is being everybody the information the world of information is really you know everywhere right now we you know use this time you know information in everyday life and everybody seems to understand this notion but it is more fundamental on the deep thing I think ok how I carry you know you know how can we codify this

1:29:25 information or some information I don't come on of course gives us a clear definition right and we are sort of it is really useful action but uh for consciousness can we use the same notion of information like channeling information as it is maybe we need other you know notions of information when we talk about consciousness that's a AIT actually that proposes actually yeah it

1:30:02 we see that I mean if but I I feel if you're proposing a general notion of information may be the the worst domain you can do is consciousness because of the empirical shakiness of the if you can you really want to propose a a theory of information maybe you should start maybe one one one way one way to start is internet but but but maybe let's agree to disagree because I'm I'm also really no expert on I think I think on this point we may have a kind of enough common ground can I be because I want to steer it back to kind of the

1:30:36 question I was originally I mean that's really interesting discussion and I don't I'm all come a resist trying to say something about what you guys were just saying but so I mean just to kind of advertise the fact that there are more than two theories of consciousness out there so global work space theory IIT there's also the higher-order theory of chirality and so the basic idea there is that it can be summed up by saying if a person is in no way aware of being in a mental state then they're not in a conscious state uh-huh so it sounds like personally you disagree with that right

1:31:12 you don't because it's not that's not part of the axioms of IAT so do you just reject I do not disagree with that point yeah but I'm saying that again that the IIT you know determines our scope of danger you know theory by actions so somehow that this the probability you are higher other aspects well it's not report ability because you could like you could not report it for whatever reason you get hit on the head in the right way and you won't be able to

1:31:44 report but that might not mean that you don't have this kind of inner awareness or something but yeah so yeah awareness of the mental functioning as a crucial element for consciousness so to me that seems to me that I mean personally I find that almost axiomatic I just don't even really a nice it means to be conscious but not in any way aware of the fact that you're conscious and I think that there's some introspective evidence that whenever I'm conscious of something that I'm also aware that I'm conscious of it so I don't know but but

1:32:16 it sounds like you're your camp you want to explicitly deny that this is a good starting I mean no no I do not explicitly denied well it's not in the accident there's no accident question a actions yeah exactly yeah what other isn't that denying that it's an important property actually it is not equivalence denying self-awareness awesome I share that yeah oh I'm sorry but I'm if so if the axioms are supposed to list like the fundamental keys of consciousness then

1:32:51 this to me seems like a fundamental attribute of consciousness yeah yeah it's not included in the axioms so that and that's one of my worries is until we can kind of decide whether or not that should be in the axioms then you telling me that IIT says this therefore conscious and I go but hey well there's still this other question about whether the axioms are missing an important property of consciousness like that it's self reflexive versus something like that which I agree would be interesting to see how you can magnetize that and whether you could have a version of IIT that had that in there is an interesting point but it so

1:33:25 that's the quote that was the problem that I was having I'm just wondering what you thought I see I see a point sir one possibility is that maybe that the only existing actions can explain the self-awareness or you know in our awareness of consciousness then that we need only the existing actions right in IIT the another possibility is that we need additional action which is like you know inner awareness that's of course another

1:34:00 possibility so there are I think two possibilities but I actually not sure which one is true of which one is you know okay well IIT so I'm not you know strongly denying the concept of in our wellness or strong that the you know essential proteome consciousness is you know in our honest actually there I mean which maybe need that additional you

1:34:35 know action yeah right okay but again that so the this is again related to the mathematics thing because that if also that the hot theory could somehow you know what's every qualifies the concept of you know inner awareness by mathematics we can sort of compare this with the notion of information the motion integrate information in EIT then we can say that we need addition action

1:35:13 or maybe this is redundant so yeah I mean so I mean I hear what you're saying it's similar as before which is that in order to test the theory you have to yeah have a mathematical a basis yeah yeah and then I just wonder why Hawk one's kind of rejoinder isn't sufficient so that there there are many ways to test a theory without having the mathematize it I mean without having mathematics is it

1:35:49 true most of our medical science hypotheses were tested without mathematics we know that the horror because we've fully worked out hydraulics we know that the Broca's area was related language not because we have any mathematical theories of that point that you know so far in our awareness is necessary for consciousness then again that the you got some empirical data brain you know activities data how can we determine

1:36:22 whether this state of brain is federal how we are in our area why is conscious based on the rocks how can we determine that without mathematics and I'm Richard answer that because I didn't bring it up well alright I would say by new interest topic hot today well it's not specifically about hot but it's about how the axioms get built and they're built in in favor of a certain theory so

1:36:58 then when that theory seems confirmed the worry that I'm having is that you haven't confirmed the theory you've confirmed your intuitions about consciousness yeah so so that that was what I mean I you know I could take or leave hot I tend to take it sometimes and leave with other trends but my question I mean I'm not trying to defend it I'm trying to oppose it I fell on yeah or yeah yeah yeah that's good yeah so I mean going you go ahead yeah yeah my proposal here is that let's have some mathematics or all of the

1:37:33 theories then we can fairly compare the different theories and notions of you know the concepts are proposing series for exam but do you think this mathematics is so important given like I mean III interested you prefer that but you kind of almost like set as a requirement for the theories is it useful I think it's very useful because of the when I bet the theory of consciousness just involves that it is sort of ambiguous to me and also I think that the in my

1:38:07 opinion you know G Global rocks face and the AIT are kind of similar because it's sample deal with the information and also global sharing it's sort of really akin to the loss of integrated information I see hey yeah this is what the interesting part is so you would think that and at some point Christophe car is in a popular article said both theories can be true but when it comes to the empirical data you see how far they diverge so Christophe has been saying

1:38:40 the NCC in high-profile places is not part of the NCC because of whatever he understood IIT in a way that is different from yours so this to me this gave me a lot of misgivings about this mathematical approach because despite the the mathematics given all the approximation everything even but then the IT camp people make very different approximations and everyone holds a mathematical high ground so this is precise but in the end they can't even agree where the central sulcus is with the textbooks and and that give me a lot of you know misgivings about this approach but again

1:39:14 that the that paper you talk about is not based on IIT or mathematics or simulations any kind of university it's a whole property of the whole research program the means have really is a research program that has been somewhat disappointing when they take that approach before they take this momentum molecular driven approach I find them more you know reasonable no no I think that the it's not to be dated you know computer simulations or anything I think just I want to compare the sort of the

1:39:47 notion of something like a notion of a global sharing or the notion of integrating information I just want to compare this notions Puri sort of motive allegory like a inverse we can say anything I think the definition was sort of maybe ambiguous to me but if you could I defy this notion with mathematics that the meaning is unique but I think I think that I think that's ultimately the point of in keeping pressing between words like you know just poetry and words and an equation

1:40:22 there is a there's a whole realm of things which science lives on which is like algorithms and and simulations most of Sciences most of a lot of actually a good part of economics a good part of this depends on that kind of language is neither poetry nor equations and they are they're precise enough to do a lot of science which is actually if you think about it when people tell you that the internet works works well in in in certain cities but not in others they are not giving you equations they

1:40:54 had some sort of quasi algorithmic and and empirical measures and say oh the bandwidth is you know this much yeah I agree I mean of course we can do many things we've dealt equations I I agree but question but not the kind of equation you're looking for there's some sort of quasi mathematical you know quantitative measures but they're usually quantitative measures I'm totally fine

1:41:27 but to me that they are real all qualitative measures put some being global box facility draw I mean nothing you can do to you know Dan and so John and Peter I think they actually run simulations to simulate a global ignition yeah yeah I think that the if there is some you know measure or growver a pair of ATL chromosome then that I say that the we can clearly

1:42:02 compare that with the notion of for example integrating I see I see you yeah Richard are you satisfied with the point oh yeah I mean I just see the debate here very interestingly as between the you know whether we start with something that's mathematical and look for the empirical thing or start with the empirical thing and try to build a kind of model of that which is the way I would say normal science kind of works but I mean that the Assumption

1:42:35 here first of all is that that consciousness is going to lend itself to mathematize ation so so your point is that well we have to start with the mathematics but you're already building in the assumption that this is something which is magnetizable but that doesn't seem to me to be obvious or warranted and needs to be argued for already before you start saying well your theory is a mathematical so it's not testable or it's not I mean so there's a worry there I think about whether we're miss applying the mathematics I see whether the property so for instance if you know

1:43:09 we're not mathematize able to me that seems to be a really fundamental element of consciousness then I would say well then all right that suggests that maybe this is out not something that is mathematize applies is how what personalizable I think for example that there are having metacognition can be mathematical right but I seems the regular people never do that right clicking and for example that in the cochrane understanding hans paper our

1:43:43 science paper that the day argue that you know for being a consciousness the system needs to have to function – what – probably why is that I you know metacognitive ability and also they otherwise global sharing a Verity so I think that here we need some you know indexes for both functions like a metacognition and also global sharing functions then we can determine the how much this system it's capable of being

1:44:18 you know having metacognition and also global I think at a very low bar you can say that there are already those quantification formatted for metacognition there is a measure of meta T Prime I'm not saying that they equivalent D prime is I think behavior measurement drive behavior you're talking about I'm talking about just so you know mechanistic you know 24 hours late this one min you can simulate an

1:44:51 agent from first principle and measure the agents better D Prime so it doesn't have to do it real work is a quantification K good is a showing how information is shareable in the Internet you can quantify it to of some sort of bandwidth measures when you create some system then you test the function of the system basically student would have behaviors it is not easier but my question here is that you're breaking up

1:45:29 a little bit can you say that again are you super Sam oh no oh no you've entered the internet they this is this is Richard pressing the button dude I have no hands-free yeah oh this is not fair which I stopped doing that are you back he deserves an answer it deserves the voice you know give back it looks like he had a chain

1:46:05 look no oh sorry you cut out for a second are you back now I'm buck I see we missed all of that you were frozen's could you I think that there is another interesting point here that you know can we clarify consciousness just intrinsically not extrinsically when you do the sound behavioral test are to the system if you can of course

1:46:39 codify the d-prime global sharing ability something just extra economy got system itself that's the only thing touching this then can we codify by consciousness it just intrinsically that's my question so I think that the theory should be able to do that like maybe that is the

1:47:13 metacognitive abilities in Poland I think that you can qualify it just by looking after the entire mechanism internal mechanisms of a system of that system yeah so this can no codification that's good I mean I yeah I think I think that's one that one we kind of have to agree to disagree some sort of approach different I think just as you can you can also study the information

1:47:46 share ability in yeah it works I mean actually I'm so I'm mentally these days I'm involved in a in the watching company the the issues that we talk about similar like do you want to talk about how fast they say all these boxing Chronicles people talk about young but how fast they can reach consensus and out and if you ask my colleagues who are engineers they we like this run let's simulate the network and then you can quantify the the number of transactions you can get is exactly like you know twenty thousand per second this is the

1:48:19 kind of measure but I can see from where like you someone like you or physicist with comfort we're coming in we said let's quantify the the network efficiency and yes my engineering colleagues were just you know to do a balk at that and said I don't but those things don't translate to the real world what good then well if you ask them up someone like my parents they would much prefer that the measures given by my engineering colleagues because that's how seemingly how how things are actually meaningful yeah yeah that's I happen in the India they arrived them

1:48:53 of course that the external part is important or relation to external world is important and some behavior measurements are not easier to interpret right and we Justin dollar measurements of network itself but uh the IITs assumption here is that only the intrinsic properties of a system determines consciousness yeah I think yeah I think this is like to me this is more way more ambitious than even string my mom I think it's more than [ __ ]

1:49:27 and strength theory that's let's agree to to you know I think I think you know it's been a very useful tool especially like I can see where you're coming from I see what how you differ from Christoph and then maybe maybe Giulio – I yeah um it's been like almost two hours awesome maybe we do have to stop and somebody not to talk to you forever yeah but you finished or over your questions you have three left but they're very quick can I just ask them

1:50:04 quick fire starters we can you share weather can you share all of that fish sounds fast like that yeah I can share them on the yeah so you can get your parallel processing yeah well let me just ask them they're very short yeah okay okay yeah so so the first one is just so I understand that um in in 93 there is a stipulation about feedback is if you let's say the the thought

1:50:41 experiment is if I can measure every neurons of the brain and then I measure them i hook up into it to it or to a network and then I make some really complicated connections and some people you may become part of my fire it might might my complex if the other complex is complicated enough understand that you then then the prediction will be ridiculous is that just by having your brain scan you might lose consciousness because by scanning your brain I hope you up your information to us some other states and at some point the whole network become

1:51:13 become the because of exclusion principle you are longer conscious so I understand that in IT through you have an axiom like you have an emulation against that which says that the basic feedback to you yes and how strong are you about that axiom at the assumption I think that this feedback is not included in the actions of variety yeah it isn't included in the codification of Phi integrate information so the before I is

1:51:49 a bow and arrow it does not take account this feedback aspects and NIIT 3.0 we includes the you know field and direction or we form a passion for codification Phi but I think that this is not that green in committed to that no I'm not satisfied with this codification actually either way yeah ICS would you would you be more willing

1:52:26 to give up the axiom give up this assumption not an axiom or give up in boxing yeah so double yeah exactly so let's say people like Specter Lama is here he would say the the feedback is extremely essential I mean that this is a miracle claim so in the I mean I actually don't always agree with the claim but even then a puker claim are you more willing to give up the theoretical assumption or you go to write off the data is what is the Wheatfield it cuz I'm shown that

1:53:03 would be quite extremely important so let's say you fit you for your own activity to my okay yeah they said I don't feed it back to you then you are still conscious so that is a good thing about IQ 3 I think oh yeah if you think I just by scanning your brain I could start yeah yeah hello yeah would you like that'll be Oh No yeah that difficult and a good question I think but I ice I think that this kind

1:53:41 of a salt experiment is you know that the stone convincing for us to assume that the you know feedback is essential for consciousness but uh see me just a new mathematical unification massive an mathematical you know that formulation of you know Phi it's not satisfactory right now and if we can solve somehow improve this I'm ok but you know in the

1:54:19 current form I think that it is just a ad hoc addition in the salary it's actually I mean our feedback easy yeah yeah so do transparency about this actually but I also view it is probably congruent with your home overall view that these are postulates so don't the fact that you are quite flexible about them and ultimately you a lot you're not like a you know militant armchair theories that actually makes me hopeful yeah but this is even more so though because the other

1:54:53 ones he doesn't think are ad hoc right the the the traditional postulates of integration caused infect power exclusion though those aren't ad hoc in the way that this feedback thing seems that Hawk is I what I heard him to say right yeah but but you know I think even for those it sounds like oh you mean sound is some more thinking them as postulates anyway which is that this one is even if used more like it if everyone is jammed it hmm don't would you agree I mean yeah again that the biggest difficulty

1:55:31 IIT now is that we count did I have the unique quantity only from actions right yes the actions cannot give us enough information for constructing theory actually right so we can have you know many versions of five many bathrooms integrate the information or information so I think that we need more constraints or the theory too you know you sharpen the you know the

1:56:05 codification was sharpen the server itself I think that it's in the current form it's still too sort of growth likes to talk to general general may I ask the second question yep yeah the second question is about the fact that they as a you know physiologist who sometime record neurons and stuff from the brain so so of course the when when you when you consciously see something and when you unconsciously see something there is a fair bit of overlap within the visual

1:56:38 cortex right it's almost as if there's the same neurons ah yeah yeah yeah yeah does that not bother you about IIT like to some extent so yeah so in the conscious case like these neurons code the same thing unconsciously is look as if it's the same neurons it's if they fire a little bit less that seems to be done and those days off for what is going on the the geometrical line is very similar between conscious and unconscious perception empirically it's just a level of activation seem very

1:57:12 different okay not I think interest question but for AIT that the conscious is determined by some states and mechanisms right physiologist neuroscientists usually think that the sort of the states correspond to some conscious facets but the iid need the causes of the sort of mechanism is also important Mecca is also important

1:57:47 for determining consciousness so that says it's very difficult sir again that the states only state do not determine the consciousness and states plus mechanisms determine consciousness so if there are two things create the same one I think that even though that you know states change it even though the state changes consciousness can be the same right yeah right okay I'm kind of

1:58:25 getting I'm running out of time I mean yeah point out which your camera is now it's like facing upward as if this is the last question actually is from my my good friend Matthias Michaud my French colleague who works on history and philosophy loss of website that he requested this question which is about I think he asks you over email about if you have a set of air molecules you know you way through the air so that

1:58:58 the air molecules that interact with one another yeah according to the theory of the axiom of exclusion at some point these is interrogatively in a way that its forms kind of network they might have a certain high Phi and I and I think it that is an empirical question right I mean itself ultimately about whether so is a variant of that question you waiting out but but you some air you move them maybe defies in the high but now if I if I just end up the question of super after question lipid si like this is not air molecules this is

1:59:31 exactly a soup a soup of some mixture of molecules yeah particular consistency so that they were built about there even like bio biomolecular engineering form to such that they would have very high five and I whisked up with a whisker that is particular bill to generate the highest you know correct opt not the highest optimal connectivity between these neurons visually these mock molecules so that they will have a – will you actually think that in that case at some point they would actually

2:00:04 be conscious so you have a soup and you whisk it in a way that an old and being power engineer to maximize Phi yeah you would think that that soup would as someone become conscious yeah critical authority on consciousness like RIT does not deal with only levels of consciousness but also yeah of course that that consciousness is totally

2:00:42 different from a lot of consciousness yeah is it a useful concept in is it what is it like is that level of consciousness a useful concept if and because these concept is I think useful for determined you know the distinguishing you know unconscious States like three from the surprise and the vegetative States from awake States I think that that these concepts is you

2:01:15 know included because of the you know to two distinct because that we want to discrete this thing we should is you know unconscious States from our awake stay ok yeah also important of course I would say this has been incredibly useful this has been way way way more useful than my trying and trying to understand the theory from from Giulio and christopher mcquarrie because they are so easy I am sure that there's no

2:01:49 like it's not like make you know explain things they also grace I to have break decay has it but but I think this is very useful and what I take home is partly your view is quite different from theirs but there is this common ground truth about the starting from mathematics and I you know mission tonight don't seem to be on the same side but I mean which and I are on the same side but in different sides and yours doors oh but I think we we can agree to disagree this has been really really useful it's fun

2:02:22 and yeah I would just say to one other interesting take home would be the usefulness of trying to develop other versions of IIT with different different combinations of axioms maybe some involving so I think the more mathematically formulate able formulated I'm sorry versions of IIT there are the better for the theory as far as connecting to the empirical data yeah that's true so I think that the I am these are some good model for you know

2:02:57 constructing theory of consciousness like in my opinion fast research starts from the definition of consciousness and like in IIT it is like an information integration the exclusion of course recap and the visa just maybe hypothesis we can call them other hypothesis but first we need to sort of declare that what consciousness is insanely like in grotto rocks Fiske theory I think it is like are they portability was sort of accessibility so first we need to

2:03:29 declare die solve the scope of the theory then we I think in my opinion we need some quantification of measurement or mathematics Asian of a theory then we can sort of discuss the difference in different theories our we can compare with the empirical data and this kind of thing I think is useful should advance the consciousness researches so I think that the IIT can be and you know defined

2:04:03 improved on also that there could be other series but I think this is sort of the right process of how you know the civil rates of consciousness contra I appreciate the approach and with produce I think when I said as in biomedical sciences we tend to be more conservative than physics but I also appreciate it diversity of approaches I think so not talking I think is always good yeah that's I kind

2:04:37 of met most of my my social media engagement provide he took the worry is that we stopped engaging then yeah and then it becomes over see it's all I think we keep talking we always find out the pros and cons I in this session I mean I I learned a lot about what what may be the potential advantages of IIT to imma say yeah so Richard as the host would you like to say something – it's the last word – shut the F up of us no I just thank you for letting me tag along

2:05:12 and watch for enemy yeah yeah should I say in the broadcast now we can still talk for a second afterwards or whatever yeah let's do these two hours is like yeah yeah