Jonathan Blow on AI Consciousness
Вставка
- Опубліковано 30 вер 2024
- Interview: • Discussion with Jonath...
Support me on Ko-fi: ko-fi.com/jona...
Jonathan Blow on AI Consciousness
Clip from Jonathan Blow
Twitch: / j_blow
UA-cam: / @jblow888
#jonathanblow #gamedev #webdevelopment #programming #ai #agi
Always fun to catch my clips in the wild! Was great having Jon on
I was gonna type out a long comment about this, but its better not too.
You know of Pinocchio, but you should play Nier Automata. And maybe go back philosophically to the topic of consciousness, and how some treat animals and how some members of certain tribes treat of people.
Consciousness is such a poorly defined term that I don't think such a phenomena exists.
This is what I tell people. It's such a useless word that it basically only ever serves to give humans a justification to inflate their ego as the "top species of the world" above everyone else
I think this is all solved if we can prove substrate independence for human consciousness. Can you scan a brain and run it? Will it act the same? We certainly either have the compute for that, or will soon, if you consider the recent Nvidia announcement. Scanning is probably decades away. But we might simulate a human-sized "brain".
It's all sound, well more fundamentally so than it is matter at least, we need look no further than 'second sound' in helium 3, as a state, for evidence of this. It is a state in which heat is observed traveling in waves; The notion of heat death, takes on another meaning when this little nugget is taken into consideration, we can no longer anthropomorphize the notion of heat dying. Our way of pushing certain rather difficult questions to the extremities of our physical understanding, into the extremely big and the extremely small.. I think though, that in a paradoxically rather similar way, consciousness, is best understood by extrapolating an understanding its behavior, by first examining it using limits, and that which those limits de facto imply, the existence of harmonics in their field.
Atoms as a model, requires that we 'flatten' or limit our thought into 1 dimension, which is far easier to do than thinking in more than our spatial dimensions, none the less, this is what we do when we conceive of them. We are transforming a 1 dimensional fact or blip on a screen into a 3d ball in our minds eye, to help us with our model of reality, which is really no more or less absurd than string theory. I digress though, to my mind neither strings nor atoms really fit that well, sound or acoustics are a far more lucid starting point for reasoning, vis harmonics.
I used to think the argument Luke gave, about how it is fundamentally impossible for a mechanistic universe made of atoms to give rise to consciousness, was compelling. You could, furthermore boil down any human intelligence to a collection of oversized mechanical switches, like seesaws in a playground, that only get activated under certain conditions. If you connect enough of those switches, would you ever have a conscious system, that is self-aware? I.e. would it be possible to design a mechanical brain, perhaps the size of a small city?
I think the whole point of emergent properties though, which was all too easily forgotten by myself, is that in spite of how unlikely it "sounds", the reality is we still don't know what consciousness is, and it still could be an emergent property. And in that case, the answer is quite straightforwardly "possibly yes". In theory if you combined enough playground seesaws together and added various mechanical cantilevers thaf flipped them in just the right way, the entire system could be viewed as conscious so long as it mimics neural function perfectly and consciousness is indeed emergent.
To refute that there has to be a stronger argument than just argument from incredulity, which is a fallacy by definition :P. By the way I love Jonothan Blow's take here. WAY simpler to follow than most philosophers, and way more logical. The fact we don't know what consciousness is, immediately destroys any arguments for dualism. It is simply absurd to even begin to start arguing for a spiritual domain, when *you don't know what or how consciousness is defined to begin with!*. That is the fundamental problem. And so long as it remains possible to define it as an emergent property (since you don't know what it is, it could be anything!), you can't rule out materialism.
Things get tricky when you try to understand causality in quantum mechanics, specifically where the wave function collapses. If we have a premise that things work mechanistically, and causality works like we expect, then we have to accept that there's a deeper process at play, that's inaccessible, which leads to the wave function collapsing. And there seems to be something weird happening when systems are observed, that needs to be accounted for when we try to reason about this process. This is a big problem in Physics, we have no real leads into what this deeper process might be, at least not any that we can be measured. So what constitutes the deeper underlying process? There could be some dynamics to the universe, which we are currently not equipped to measure and study objectively, that play a pivotal role in the appearance of the complex consciousness we experience. That are deeply intertwined with the dynamics that produce the physical world.
For me, thinking about the universe and experience as originating from the same deep fundamental process is deeply satisfying, and overcomes the dualism problem, even though there's no real objective experiment that we have that can confirm this. Saying that consciousness is a layer above something fundamental, governed by the interactions of particles let's say, is a very disheartening way of looking at the world, is not any more supported than the previous view, and leads to a ton of conceptual loose ends when we try to think about causality at the deepest levels of the observable universe.
I guess, in short, seeing consciousness or consciousness-like phenomena, as being vital to the movement of the universe, can provide a deeply satisfying resolution to a lot of these intellectual problems when we reason about consciousness.
David Bohm has done a lot of work in this area, physicists like Basil Hiley are continuing his work. Work in the philosophy of science by David Chalmers and contemporaries explores ideas of "proto-consciousness". All trying to resolve problems of causality in QM, and accounting for the appearing of our subjective reality.
The thing is - if there’s a word for consciousness then there has to be at least some meaning to it. This is obvious, but shows that we do know kinda what we mean by consciousness, even if it’s not put into words precisely.
It's information processing
pov: bro science. "particles dont exist, only waves"
I don't know how much Jon is into philosophy, it sounds like he's decently familiar with the philosophy of mind, but I think he would get a lot out of reading Galen Strawson.
The Eliza effect will end mankind
"What is the file size of a real memory?" is quantifiably impossible and qualifiably defunct hypothesis. But one worthy of asking nowadays due to AI. So that WE don't think OURSELVES to be like stochastic parrots. At what level do we apply boolean logic to our "circuitry"? At the synaptic level? When a particle gets/looses a charge? QT field excitation? Virtual particle collisions?(those too, aren't irrelevant to the whole) And how would we then define the gate's I/0? Per cubic planck lenght (smallest measure) in some lattice for example?
Even if you answer this, our memories stll aren't the same twice, unlike those in a hard drive. So you have to account for the continuously changing properties of the organism. Which then forces you to consider chemicals coming from other parts of the organism, required for its functioning, and how those affect the brain (would the work of proteins also be given their 0/1 values?). Which then, since those are ever-changing too, you have to consider every aspect from outside of the organism. As the elements from the environment you come to contact with also affect your internality (about every 7 years, all cells in your body are not the same, etc). So an organism, functioning here as the plane of emanence for whatever "consciousness" is, is more of a non-fungible pattern from which matter passes through (is assimilated, dissimilated).
Strings of binary ended up having their meaning derived from agreed upon conclusions by humans, they were not immergent. They rely on an interpreter capable of abstract thinking in order to have properties at all.
Hence, even if you could show me a perfect simulacra of myself in a digital medium, I'd still say whatever mollusk is in my nearest vicinity to be closer to "consciousness" than my copy. Because it too is composed of these non-fungible (I'm sorry for using this term) patterns. My copy is then a perfect p-zombie, but it is very convicing! And unfortunately, that is enought to change the world. The question that matters, then, is an ethical one. "In what type of world do I live? What is likely to be done to me by systems of power now that there is a convincing enought double? Am I to be exploited, discarted, etc".
Honestly, if this AI revolution is going to be good for anyone, it won't be for people with our sensitivities. The ramifications are too antithetical to what humanity has been thus far. It's hard to have relationships but we muster through. It is hard to learn a craft but we muster through... If there is a "easy-mode" to all these (especially if it's considered more "productive"), then such a facilitism will be taken 99.9% of the time. The path of least resistence will make it so the future is no longer in our hands, as we outsource too much of ourselves to AI's reliant on infrastructure owned by a few technocrats. It'll require a lot of social enginnering to get to the point which the newly arisen "useless class" can in anyway be made to match the values of the owners of the infrastructure which they rely upon.
I can choose to keep in mind Aristotle and St. Thomas, who explain all this stuff neatly and consistently. Or I can put them aside and try to make sense of what remains. And what remains is that everything turns into an unknowable appearance, and then you can have software or robots that can theoretically be made to appear like conscious, thinking and feeling beings. And who can theoretically fool everyone, but practically only people who don't understand how these programs work.
What if I say to you that human beings are deceiving you into thinking that they really are conscious and self aware? What If we were only a biochemical machine, programmed to feel and think because these traits are useful for survival?
@@lucarossi8442
["What If we were only a biochemical machine, programmed to feel and think because these traits are useful for survival?"]
I'm 100% sure I'm conscious, it does not matter if it is or not because "these traits are useful for survival", in a naturalistic world it would be the case and still it would not allow people to dismiss consciousness.
The loss of metaphysics in education and intellectual circles has made this contemporary existence excruciatingly tedious.
7:30 "there must be something else". Maybe not. That might be wishful thinking.
If it (Devin) can't even make a secure React site, why worry about AI Consciousness?
Category error.
It reminds me of a lecture one of the philosophy professors gave at my Uni. It came down to him arguing that ticks only having some sort of "consciousness" when there is a warm blooded creature nearby. When they're waiting for their food on a leaf, nothing really happens inside them. They are in stasis. When a warm-blooded creature walks by is when they awake from stasis and there is some form of processing going on inside a tick.
Similarly, LLMS, are in stasis most of the time. If they can be conscious, they can only be conscious while they are providing output to a user. Interestingly, if it were aware if this, it would bias LLMs to longer and longer answers as to not lose consciousness. It definitely a strange twilight zone that deserves some thought.
@@saltyscientist1596, when it comes to AI, people mostly use their imagination and then create theories from that, as if they were talking about some observation or real possibility. Maybe two percent of people who talk about AI have any idea how it actually works and therefore what are its limitations. Other than that, this is like speculating whether a vegetable soup will become alive and walk out of the pot if you stir it long enough.
I found this somewhat incoherent but somehow managed to enjoy it anyway.
Confirmed: Jon Blow subscribes to the Everett interpretation of QM.
6:47 And no, I don't think Jon readed Searle with that take, that assumption has almost literally nothing to do with his argument.
At this point I won’t be surprised if this video was generated by AI.
2:39
Nah, this is just a very bizarre mistake of thinking that by don't do deep thinking it should mean it is "automatic", and things don't work that way. Volition is an important part of our behavior, and volition itself is conscious and directed by the consciousness, the act of deep thinking is only possible because we have these immediate volitions that are fast and can deliberate very fastly about everything, I never even by a single time felt the difference between deep thinking and fast thinking, it is just me acting, literally, it is conscious self and I cannot remove that conscious component from every single directed action I do. The problem here is that you are seeing a disconnection between your self at different moments, so you maybe really need to see a psychiatrist, just saying.
None of what you said has anything to do or grounded within actual neuroscience and psycology of human beings.
He says we have automatic behaviours because we _know_ from experiment we have behaviours so automatic that you can trick actual human beings into remembering, moving muscles carrying out actions _they don't even remember_ or wernt' aware they did. That is fully automatic behaviour. If you go a step furhur we have reflex nerve pathways that quite literally physically bypass the brain and carry out actions. That is the opposite of conscious
Very rude to say "you need to see a psychiatrist" just because you didn't bother to study any human neuroscience.
@@Mallchad
You say: ["None of what you said has anything to do or grounded within actual neuroscience and psycology of human beings."], and I say: proofs?
You simply don't know what are you talking about, and you did not understood anything I said in this comment.
The evidence: ["He says we have automatic behaviours because we know from experiment we have behaviours so automatic that you can trick actual human beings into remembering, moving muscles carrying out actions they don't even remember or wernt' aware they did. That is fully automatic behaviour. If you go a step furhur we have reflex nerve pathways that quite literally physically bypass the brain and carry out actions. That is the opposite of conscious"], see?
You don't know what are you talking about. When, I repeat, when I said anything about human beings not having automatic behaviors? I don't recall it, I did not said it, you assumed it because you (wrongly) assumed that I am not aware of these things and that I am not aware of these discussions and positions, well, I do. I do know we have automatic behaviors, I do know we have instints, I also do know and understand that not everything in human behavior is intentional, and this have nothing to do with what I said in my original comment, dot.
I said specifically that the opposite of is not , this is not stablished neither in neuroscience nor in anything. I'm not saying that there are that are also , rather, I'm saying that it is not true that because you are not doing deep thinking it means you are being on "automatic mode", this is what we call a straw man, a comprehensible mistake to make but not so cool if intentional. You just need to know transitively that if deep thinking is made of parts, these parts must not be deep thinking itself, they should be volitions and shallow thinking, and if all of this is then by extension deep thinking would also be (a position I don't think even Jon is holding here, because it entails profound conclusions; conclusions that are not the default neither in neuroscience nor in phillosophy of mind). I'm saying that if he thinks that, while he is speaking with someone, most of the time it is not him but some automatic and external thing, this sounds for me a psychological problem more than anything (maybe even a neurological condition), it does not sound normal that you feel desimbodied of your own volitive actions and free will, this is the catch on my comment. I'm not trying to be "rude" (even tho this is not a problem on itself, Jonathan himself is rude with things many times), I'm really worried about his mental wellbeing, because this do not appear to be normal, dot.
Any more thoughts on me "not bothering to study any human neuronscience"?
@@diadetediotedio6918 I think you are underestimating how often automatic behaviours are and it extends as far as automatic thought patterns.
You don't have to be actively thinking about things or have the capacity to perceive or experience those thoughts, or muscle movements for them to happen
Yes I think we can have "automatic" deep thoughts that is my thought not Jons. I would suggest it is backed in neuroscience because we have things like split brain experiment and experiments with subliminal messaging that very clearly points towards a deeper processing that you aren't aware is taking place. As well as being asked to remember things and having the subjects quitely edit the memory before it is recalled unconsciously and out of their control
This is what I believe Jon is has realized whilst struggling to put it into words and why I think its "rude" to essentially dismiss it as just a "problem" with him. I don't believe much of what you said addresses this _except for_ the mention that it would be weird I suggested there is automatic deeper though
Another way to think about it in some traditions is that the external world is inside of consciousness rather than the other way around. The world being a sort of mental construct, under this paradigm.
Are you saying this in my consciousness or your consciousness?
That's the hermetic conceptualization of reality: that a being for which nothing can exist outside of itself, has no raw materials with which to create anything and so, the only way anybody or anything can create something without external materials is with thought. You can picture all sorts of things in your head.
@@seriouscat2231 why not both? A single consciousness could play a game where it sort of fragment itself and plays different characters.
Either way, what we consider the world could be a mental construct created by the interaction of many smaller consciousness bits. In some traditions everything is conscious, so you could say that the world is made up of all of us perceiving each other in different ways, but without an external reality underlying it all. The perception is the thing that exists. That would probably match what we know about certain phenomena observed in the quantum world.
Though of course there are many regularities in our perception of the world (gravity, for example) among all the different actors, so you would still need to explain that somehow. I can think of a few ways to do that.
@@aazendude true, I remember a near-death experiencer mentioning our world looked like "fragmented consciousness" from his perspective, like shattered glass. Meaning perhaps that what would be perceived as a unified field of consciousness usually, was perceived here as a myriad of independent actors. Some spiritual traditions include the concept of the "illusion of separation" and the fundamental oneness of everything. But idk, it's mostly theoretical for me since I haven't gone through the spiritual awakening some people describe which apparently gives you insight into this matter. I enjoy thinking about it.
@@Jaime-eg4eb, is there any reason to believe that those "some traditions" know anything about reality?
Agreeing with jon, once more
consciousness just evolved, didn't it? because it was advantageous in tribes to have empathy or theory of mind, as it leads to more collaboration. so the tribes that contained more self awareness led to increased prosperity. isn't it that simple?
Consciousness is an illusion, whether in a human or machine. It's just a matter of how believable that illusion is.
What is it then that is conscious of that illusion?
@@seriouscat2231 For now, humans. Later, probably machines as well.
I guess you're trying to find a contradiction in my words, and I guess you don't have the contradiction you think you have. But I'll be glad to be surprised for either guess.
@@mohamaddelkhah, simply, how can an illusion be conscious of anything?
@@seriouscat2231Humans aren't an illusion. Their consciousness is. Humans are conscious of the illusion of their consciousness. So this isn't a case of "illusion being conscious of something". (Not that it would pose any issue even if it was, since there is no discovered law of nature against illusions being conscious of something.)
Also "being conscious of something" doesn't imply having "non-illusion consciousness" on its own, since it's compatible with having illusory consciousness as well as real consciousness.
It's obviously not an illusion, determining whether or not something is real or illusory requires consciousness in the first place, claiming that consciousness is an illusion presupposes that you are conscious enough to make that claim. Like all post-modern attempts at deconstructing reality, it only deconstructs itself.
For me AI trully consciousness, whenever its can rewrite and compile own code.
What a low bar to set
@@josephp.3341 i mean AI can rewrite itself
Arguably, someone afflicted by dementia is no longer capable of updating their own code; do they then cease to be conscious?
@@orlovskyconsultinggbr2849, nobody who knows how it works believes that.
What if its written in an interpreted language?
Either: 1. Physicalism (in a perhaps overly broad sense; you hopefully know what I mean: include pancomputational theories here as well) is true and consciousness is an example of a potentially strongly emergent phenomenon (also, I'm including consciousness "as if", such as e.g., consciousness as imagined by Joscha Bach, Douglas Hofstadter, etc. in this bucket - I really dislike when they use the word "illusion", but in this context that's an irrelevant semantical point). Or 2. Consciousness is itself a fundamental force; note that something is fundamental when it cannot be explained as being the product of interaction of other tings (forces, particles, whatever). To me the latter option is the overly complex one: as here the material/mechanical part of reality needs* to be organized in a particular way for the then supposedly fundamental consciousness force to then manifest (rather than emerge). A fair few have accepted a mechanical explanation of life, I think that like in the case of evolution, it is the human need to feel special coupled with their spiritual feelings/bias that prevents them accepting that there is nothing mysterious about viewing consciousness as the result of interacting entities. On that note, we also still don't have a good general understanding of what life is (yes, the freshman biology book on my shelf - IIRC - defines life in terms of the cell, but I hope you can see what this isn't exactly satisfactory; perhaps if you generalize the cell and end up with some agent like entity whose actual physical boundaries aren't as important as its informational ones.. well, that's a tangent). The point is, we also don't need to fully understand what consciousness is - and we also do not know what reality fundamentally is, so we don't really know what anything really is. But we can start somewhere; and since one of the only things we do actually know is that reality at the very least supports computation, then a computational theory (of life, the mind, reality) seems like the best thing to explore first, then, if we are wrong, we will at least actually understand why that is so (as opposed to how it is now where we just jump straight into making wild assumptions about the existence of continua, abstracta, mysterious elements of conscious experience, etc. - all theories, which, btw. still needs to be translated into something computable assuming they are supposed to be able to produce predictions; integrals are necessarily translated into sums and fluffy words into logic, etc.). Or 3. ???? :)
Footnote:
* "needs to be ..."; based on our current understanding. (Even in panpsychist theories the organization of other entities matter; and in any case we don't have commonly accepted evidence to support the contrary).
Wrt. what he says at the beginning of this clip: I hope he is right. But I fear people will instead be overly protective of their ego and hence never accept anything not human as conscious. Rather than risk being wrong (and hurt their fragile ego by looking stupid) they'd rather risk causing harm. And I think they'll rarely - if ever - even approach being conscious of this. I think this is what we are already seeing now.
So we should shame people into believing it is conscious, whether it is or not?
Who gives a crap about ego? You’ll know if AI has consciousness because there are markers for that we subconsciously look for. It’s not a matter of believing. Do you just believe and not know that people around you are conscious beings?
@@seriouscat2231 Should we shame people who kick dogs thinking they don't have the capacity to experience the negative emotions associated with pain? This is in fact what supposedly civilized European people used to believe until quite recently. I don't know exactly how it changed, but shaming might have played a role in it. Just so you don't misinterpret me: this is NOT me saying "yes, yes, we should". This is a complex issue; but I'd rather see us err on the side of feeling dumb rather than risk causing harm.
@@WhoisTheOtherVindAzz Dogs have a biological nervous system extremely similair to ours. We are both carbon based like all other life. We know what pain is and feels like. Not the same can be said to a digital system. Especially if accepting these "systems" as "maybe kinda conscious" will raze the livelyhoods of countless knowingly conscious beings for the good of technocrat's profits.