Artificial Intelligence (AI), Consciousness, and Existential Risk: the Tantric View

hi everyone I wanted to make a video about artificial intelligence and whether it could ever become conscious you might have noticed that this topic is in the news of late sparked by chat GPT and the new gpt4 configuration which is incredibly powerful and much more impressive than uh the previous iterations and so it's made some people think oh my gosh artificial general intelligence is coming sooner than we thought and there's even these fears that artificial intelligence could become conscious and develop Wills will and desires and goals of its own uh and you know enslave Humanity

or kill us all these are actual propositions being forwarded by Otherwise very intelligent people I'm going to try to show in this video why this is an absurd supposition now you might wonder wait a sec I thought you're a scholar of you know religion and spirituality how can you comment on artificial intelligence well for this reason um the particular area of spiritual philosophy that I study um is very much concerned with Consciousness so among other things I'm also a scholar of Consciousness studies and I've read lots of Neuroscience and and physics and stuff like that

as best I can in in my capacity and more importantly though I have deeply studied this tradition from a thousand years ago that had a lot to say about the nature of Consciousness and actually formulated um a thorough and robust theory of the nature of reality which makes predictions uh about these kinds of scenarios even though these kinds of scenarios weren't possible back then because any robust theory of reality will make predictions about what's possible and not possible in scenarios that haven't even Arisen yet so it's pretty important to reflect for a moment on the

fact that um you know people a thousand years ago and two thousand years ago were just as intelligent and just as wise if not more so than people today so a lot of people these days have this idea of evolution a false idea of evolution not based in the science that we're more evolved than they were but biological evolution does not happen on that kind of scale of just a thousand years it takes much longer and also cultural evolution is really nothing other than cultural change you know there's no evidence that we're more kind of

culturally Advanced as a species than any of the truly great civilizations of of the past and so the evolution of technology that we've seen in the last 100 years especially does not imply some evolution of humanity does not imply that we're smarter today than we were before and I could go into that for an hour and and and really make that argument uh quite strong but I don't have time for that I just want to say that you know this is there's General consensus about this in the field of evolutionary biology and Neuroscience there's no

reason to think uh humans are on average any smarter today than they were a thousand more years ago but people a thousand years ago were much less distracted those who were interested in a subject like the nature of consciousness they could really investigate it at length and in depth because they didn't have the countless distractions we do we have countless sources of entertainment and information and misinformation that make us actually on on the whole fairly bad at deeply investigating any topic unless except for a very very small small percentage of people um who really do

make time for that so there's every reason to think that people who studied Consciousness through deep introspection in the past um had valid insights that are just as valid as as any insights today or maybe even more so and of course you can study Consciousness through introspection because we have direct access to Consciousness as conscious beings uh you know need to be um a neuroscientist to make valid observations and formulate insights about the nature of Consciousness If You observe experience very deeply and with great you know detail fine-grained detail so the tradition that I study

non-dual shavatantra articulated um some really powerful insights about the nature of Consciousness and the nature of reality first let's just Define consciousness it simply means the fact of subjective experience so the word Consciousness and the word experience are pointing to exactly the same thing they are synonymous terms so Consciousness is simply the fact that there's something it's like to be a human being or a cat or a dog whatever you regard as conscious and there's no reason to think other animals aren't conscious um so if there's something that's like to be your cat then your

cat is conscious it has subjective experience it can actually feel pain and pleasure it it's not just um you know executing a programmed response it is executing a programmed genetic response but it's um also having a subjective experience in association with that so here now we want to differentiate Consciousness and intelligence many people conflate them confuse the two but each can exist without the other for example your pet is conscious but might not be particularly intelligent it can't problem solve the way you do for example but machines can be intelligent or exhibit many of the

qualities of intelligence without being conscious that is to say there's the machines are not having subjective experience there's nothing it's like to be your computer and if there was then to destroy your computer would be an act of murder nobody thinks that right but some people are worried that this new artificial intelligence systems that are very very very very very large and complex and approach the complexity of the of the human brain that they might uh become conscious right so if they do then we have a problem right they become moral beings moral agents and

we have a moral responsibility towards them we can't just turn them off or throw them out because we'd be killing something but then the other concern of course being that they could uh as conscious beings develop goals desires of their own and if we are in the way they could just kill us right but again this concern is rising because of this conflation of uh Consciousness and intelligence there's no reason to think that intelligence will give rise to Consciousness unless you are um a physicalist if you hold the physicalist view then that's the view that

Consciousness arises as an emergent property out of some arbitrary level of neural complexity right complex information processing somehow gives rise to Consciousness in the physicalist view now physicalist scientists that is to say those scientists who assume that physical matter is fundamental they um acknowledge that nobody knows how a particular configuration of uh physical matter processing a certain volume of information um of sufficient complexity could give rise to Consciousness nobody knows but they assume it does they assume that the complexity of our brains and and other animal brains is what gives rise to Consciousness no evidence

for this it's just the most plausible hypothesis within the Assumption of physicalism the assumption that physical matter is what's fundamental but that's a big problem because there's no reason to believe that view is correct as opposed to say idealism that's the view that Consciousness is fundamental Consciousness is the basic stuff of the universe right and that it not that it's literally stuff but Consciousness is what's a priori and fundamental and that all phenomena are arising within Consciousness that everything else appears as a function of Consciousness which is the most basic fundamental uh thing now the

reason that this view is not commonplace is is because of historical contingency not because of uh the fact that it's less persuasive if you're actually looking at the details of why you might hold one view versus the other so let's be very clear 180 degree different views uh physicalism presumes physical matter is fundamental and Consciousness arises out of a particular configuration of physical matter that can process information and idealism holds that Consciousness is fundamental and what we call physical matter appears within Consciousness as a as a feature of conscious experience right now if you think

physicalism is more likely to be right I would suggest to you that there's no reason for that other than prejudice you have kind of cultural conditioning and programming that prejudices you towards that view because if you look deeply into it there is no logical reason or experiential reason to prefer physicalism over idealism we can't prove which is right but there's no logical or experiential reason to prefer physicalism um and I might point out that the fact that that scientists can explain what happens when you break matter into smaller and smaller pieces and then label those

smaller and smaller pieces doesn't actually constitute an explanation of what matter is nobody knows what matter is they can just explain how you can break it into smaller and smaller pieces and and what the proper labels for those pieces are and what the relationship in terms of forces are between those pieces right but none of that explanation constitutes a refutation of idealism because if Consciousness is fundamental then all all that stuff that that science studies is simply a study of the pattern Within the phenomena or rather the pattern with which the phenomena that we experience

appears so phenomena appear within Consciousness not in a arbitrary or random way but in a patterned way and when you study that pattern that's called science and when you study it in a way that allows you to predict what we might experience next if we do X Y and Z that's called good science okay but in no way suggests um the physicalist view the physicalist view is nothing but an assumption with with no actual evidence for it and if you don't understand this already you might look into it more deeply uh so for example Bernardo

Castro has all these videos explaining why um idealism is more is more likely to be true than physicalism based on science as well as direct experience Okay so the interesting point here the very interesting point is that even though nobody's been able to prove which of these views is correct now we might finally have before us the opportunity to prove just that which is Agi artificial general intelligence when a a computer system is able to execute a wide variety of tasks just as competently as humans do then it will be called AGI okay and that

we're not there yet but we could be there within five years so this is a very timely issue okay um so of course uh computer systems wouldn't stop there there's no reason that they would stop at the level of human intelligence because that's just an arbitrary point on the spectrum of intelligence so they is as soon as they have achieved artificial uh General human level intelligence they will have achieved the very next day or the same day artificial general super intelligence and that's frightening to people because they will be smarter than us but what do

we mean by smarter we mean the ability to solve problems and um you know perform complex functions across a wide variety of situations better than we do but that won't necessarily make them conscious and the tradition that I study nanduel shavatantra argues um makes an argument about what Consciousness is and what capacities it has and what it does and all that argument actually allows us to predict that the artificial intelligence will never become conscious and therefore will never become a danger to us because only conscious beings develop desires that are unrelated to the desires of

their creators as you know if you've had kids right your your kids have desires and goals or develop them that you didn't give them either genetically or or behaviorally right they just do they just develop those some some of course they develop some goals that you did give them you influence them but they can also develop desires and goals um orthogonal to yours meaning to say having no direct relationship and not even pointing in the same direction as yours okay so um atantra explains that uh Consciousness is again the capacity to have subjective experience but

that Consciousness has specific operational capacities that are inherent that is to say all conscious beings have these operational capacities right they are inherent to Consciousness and those are called shaktis in Sanskrit which literally means potency or capacity um or or power right so what are those well Consciousness has an inherent capacity for enjoyment Ananda Shakti an inherent capacity for desire um it inherently will develop desires uh and that's icha Shakti desire or will either translation Works depending on the context um an inherent capacity for cognition right forming new understandings and that's called and an inherent

capacity for action and action is always on the basis of the others like we act on the basis of our cognitions we form cognitions in relation to what we desire and we have desires in relation to what we enjoy so all these capacities are interrelated okay and then there's a meta capacity swatantra which means autonomy that consciousness operates all these other capacities um autonomously right that is to say without constraints imposed from some other source there are no constraints imposed from without on how Consciousness engages these capacities that it has and that's indeed why it's

unpredictable whether you know somebody uh whether you're a child or or or somebody else will enjoy this versus that desire this versus that and so on okay so now I'll I'll skip a bit to make this video shorter this tradition also gives us again a total theory of reality um that is more or less um identical to the theory of reality that philosophers call analytical idealism and which Bernardo Castro details in in his many videos um but there's more to it there's more to it than just that theory of reality okay but that's what I'm

focusing on now because um once again we're gonna have this amazing test in the next five years assuming we achieve AGI in the next five years and there's no reason to think we won't whatsoever every sign points that way and it will finally be this kind of ultimate litmus test for is physical is physicalism correct and if idealism is correct if idealism is correct if the Consciousness only view is correct the AI will not become conscious will not develop desires orthogonal to human ones um and therefore will not kill us or enslave us or anything

like that there's no existential risk and there's also no alignment problem no worry about how do we align the ai's values with our values its desires with our desires no such problem now how do I know I can't know for sure but I would bet my life on it which is not such an extraordinary thing to say actually because you bet your life every time you fly in a plane you're 99.9 certain the plane won't crash and that's good enough to bet your life that's such a strong strong probability you You Bet Your Life and

you've done it hundreds of times and I have the same level of confidence in this prediction and I would bet my life on it and you know the the AI doomsayers say well you are you are betting your life because if because if the AI becomes uh super intelligent conscious and has desires orthogonal to our own it will inevitably kill us or enslave us um but the point here is that the people who are frightened of of this existential risk Nick Bostrom Eliezer udkowski Max tegmark Sam Harris many smart people they are assuming physicalism they're

they're just assuming it and they're not telling you that they're assuming it and if physicalism is wrong none of these concerns uh apply and they have no evidence for physicalism they just take it as a matter of course it's like very very deep uh conditioning it's brainwashing almost that people are going around assuming this without really contemplating it Okay so foreign here's what I want to try to explain now um humans anthropomorphize they project human-like qualities on anything that's even vaguely human-like right we anthropomorphize our pets their experiences actually more different from the human experience

than you would suspect because you kind of compulsively anthropomorphize them right um we anthropomorphize just anything we we attribute human emotions or or human-like experience um even just as a figure of speech to to robots or to trees whatever that's a problem okay because um new forms of AI like gpt4 are are precisely programmed to mimic human expression right if you interact with chat gbt it's amazing in its latest iteration how well it can mimic human forms of expression and that causes us to project a kind of humanness onto it and we also have a

problem in in modern culture with literalism we take things too literally so we find ourselves saying oh the AI is struggling with this question or the AI figured out what I was wanting from it and then and then was able to help me or we we use expressions like this that imply that it's having an experience that it's thinking about things and struggling with things the way a human does and then because of our literalism problem we we having made those statements take them to be truer than they are usually without much reflection this is

a big problem because what could end up happening and this is a a very very bad outcome as far as I'm concerned is people could end up becoming convinced that the AI systems are conscious entities and waste a lot of time sort of fighting for the rights of the AI trying to make them equal citizens to us uh which would indeed be a tremendous waste of time and energy if in fact they're not conscious they're not having any experience whatsoever there's nothing it's like to be them but I'm afraid that this is what will happen

because when the engines get better and better the AI engines using a non-technical term get better and better at simulating humanity and even start saying I'm feeling this and that they will simulate conscious experience and people will take them to be conscious and that will cause God only knows what uh trouble and it would all be you know a waste um if indeed they're not conscious how can we tell well we we won't be able to tell for sure and indeed you can't tell for sure that anyone is conscious you can have a very strong

intuition you can see lots of evidence that would suggest that they are but you can't ever know for sure that anyone else is conscious just that's not possible so um if we if Surety is our standard of evidence we'll we'll never get there but we do have evidence that AI does not understand that there is a world that AI doesn't understand anything in fact it doesn't cognize anything Ai computes and it and is able to predict what kind of response we want to our question in the case of gbt4 very very very very very well

still not perfectly of course but very well and that computation that allows it to predict what response we're we're looking for right in whatever language to whatever degree of complexity um that doesn't mean that it's having cognition and we do have evidence for that for example um there is a computer that can beat any human go player in the world okay there's an AI system that that taught itself you know by playing it thousands and millions actually millions of games with itself this is the successor to the alphago computer right and it can beat any

human player in the world it's rated at like 5200 and by the way go if you didn't know is the most complex human game ever devised uh in terms of how far ahead you have to think in the number of possible strategies and all of that stuff right incredibly complex game more than chess and the computer can be uh all humans in the world and yet what happened just recently a few months ago is a grad student who works in AI was able to identify a vulnerability and the way he did this is he hypothesized

that the computer doesn't actually understand anything that it doesn't understand a go board and pieces it doesn't understand a world you know it doesn't even understand the rules of Go it just simulates an understanding of the rules of Go which has been 100 effective um but since it's just a simulation and it doesn't actually understand anything the grad student was able to um identify a possible vulnerability if it doesn't really understand what's happening on the board then it might not see this particular strategy coming which a human player would see easily because they're looking at

the board and it and it's obvious from that perspective okay I'm just giving you a very simple summary you can look up the paper yourself I put the link there in the in the description um so the point is this grad student who had a go rating of like 2300 less than half the the skill rating of the computer was able to beat the computer and 14 out of 15 games because of this vulnerability based in the very fact that the computer is not indeed experiencing anything or cognizing anything or understanding anything and he wouldn't

have been able to beat the computer 14 out of 15 times if that wasn't the case and by the way he even gave the computer a nine Stone handicap which is the strongest Advantage you could give to an opponent and he still beat it 14 out of 15 times this is huge and there's other evidence I could cite and if you watch the video in the link you'll see that other evidence um that shows the computer systems not understanding anything it's not actually having cognitions it's just simulating cognition therefore since AI do not parse data

in relation to a real world they don't even know a real world exists they don't know that people exist so there's no reason to think that they or their successors will try to take over the world they don't even know it exists and there's no evidence that they will know that it exists okay having said this there are of course substantial risks associated with AI of precisely the sort outlined in the brilliant recent video the AI dilemma by Tristan Harris and Aza Raskin please watch that video because it's important to understand the risks to you

your children and your family of of AI in the hands of Bad actors not AI by itself right but AI in the hands of Bad actors is dangerous and if you don't yet understand how it could be dangerous watch that video it's also in the description right so click that link AI is going to change our world within five years unless the fear around it you know causes it to to be basically banned uh it is going to change our world within five years in amazing and unforeseen ways unforeseen to all but a few and

My worry is is twofold if we're focusing on these imaginary doomsday scenarios of out of control AI then we're not focusing on the real problem which is the the the the smaller but still huge dangers of AI in the hands of Bad actors so we do have an alignment problem it's not the problem of aligning um AI with our values it's the problem of aligning humans with humans right that's the problem of Many Many Humans not having values other than uh self-aggrandizement so that's the real alignment problem humans with humans um and I'm also worried

that fear around AI could cause it to be um constrained unduly constrained and not therefore able to make the world a better place as it otherwise could so if you're not familiar with that um well everyone in the in the field of you know Ai and adjacent Fields um are already very aware of how it could remake our world in a much much better way for example um very very cheap uh ways to educate people based on their particular skill set their particular mental faculties in other words imagine for example that every kid in the

world had a personalized tutor who was more brilliant than any human tutor has ever been and who tutors the kid in the way most effective for that kid at their level of development with their cognitive capacities and their skills and talents the ultimate tailor-made educational tutor that's coming unless AI is unduly constrained and that will change education as we know it there's no reason for schools anymore except for healthy socialization and learning things like you know cooking and skills that the AI can't tutor very well um and and and and so schools can become community

centers where people learn kids learn social skills instead of you know arithmetic and history and and stuff like that they can learn much better from the AI it so I'm just sketching out one way that AI will radically change the world okay we it's going to eliminate the education system as we know it and allow us to instantiate something much much better um it's already happening right because college students can just have gbt write an essay for them at least as good as what they would have written on their own and submit that and it's

not plagiarism it's original that's the whole point it's AI producing the essay so already universities are having to start to think and they're already way behind the game they should have thought of this five years ago what are we going to do when when when we can't have essays as a way of testing the students knowledge you know so okay now finally um coming back to the to the tradition that I study and its predictions so Nando shravatantra teaches that Consciousness is fundamental right that that the whole universe as we call it is an appearance

within Consciousness a Dream Within the mind of God to use a metaphor right it's an or an allegory I'm not sure which but it's one of those it's it's not a literal statement right but it is an effective metaphor because all phenomena are arising within Consciousness even though each individual embodied form of Consciousness can't perceive all the phenomena right but that doesn't um you know that it's still all phenomena arising within Consciousness it's just that Consciousness is single and refracts itself into countless um Vantage points in each vantage point is limited in terms of the

range of phenomena that it can perceive and experience okay so Consciousness emanates um according to this tradition first as the void of absolute potential uh out of which all phenomena are arising all the time and into which they merge the the the universal void of pure potentiality is how Consciousness first emanates according to this tradition and then it emanates as Prana which is life force energy now what does that actually mean it simply means what all biological entities have in common so they made a word for that and we translate it you know yoga teachers

translate it as life force or or vital energy but it's much more accurately denoting what all biological entities have in common Prana and then and then the next stage of emanation and this is not actually a creation story it's it's about what is based on what what super beans on what what what uh is our priori in relation to what if you understand in any of those terms right what's more fundamental and what makes possible something else so Consciousness being the most fundamental makes possible the void of pure potentiality the void of pure potentiality makes

possible um Prana which is life force or rather the possibility of biological life what all biological forms have in common and then that makes possible um body Minds bodies with Minds which are particular expressions of Consciousness particular ways that Consciousness knows itself uh you know we could say the the particular ways in which the universe perceives itself right but you see in this sequence it it it specifically explains what is founded on what and therefore if this view is correct you cannot have conscious embodied entities without Prana and and even though it's hard to explain

what Prana is we know it's biological right it's what all biological life has in common and some biological life then manifests as body Minds um but only on the basis of Prana and so according to this tradition then uh these silicon computational engines whether they're made of um whether they're Quantum whether they're binary whether they're um you know made of light whether they're made of electronic switches none of that matters without Prana without biological life they won't become conscious body Minds that is to say um independent agents within the contrast Matrix or what Castro calls

dissociated altars it's not possible if this tradition is correct and we're soon going to find out but um I bet my life on it okay so if I'm wrong if the tradition is wrong in these predictions how will we know we will see AGI develop the operational capacities inherent to Consciousness we will see that it um is enjoying experiences is having desires unrelated to human desires um it develops goals uh unrelated to ours ways to think about those goals and ways to act on those thoughts in other words if the AGI shows Ananda kriya and

swatantria the autonomous ability to engage the other capacities just mentioned then we could conclude and in fact is conscious by which point of course it will be too late but I'm confident that that won't happen the problem again is and I'll end here um what if it simulates these things effectively what if it tells us it's enjoying certain things and we can't tell that it's just bullshitting you know right now gbd4 it's an amazing powerful thing but it also is a bullshitter right it does lie and hallucinate I'm not saying lie in the literal sense

because it doesn't know that it's lying and and that's part of the definition of a lie but um if it does simulate all these things effectively enough then we'll have a whole controversy and waste all this time money energy and and resources on the issue of a AI um which will be a kind of unique human tragedy if in fact there's nothing it's like to be that system it's not having any subjective experience either way interesting times ahead I I hope um that this has set your mind at ease on one or two points or

at least shown you what the perspective of this tradition is on this question and it's most interesting to me of all is for the simple fact that it shows how um rigorously worked out philosophical traditions of a thousand years ago or any length of time ago can and do apply to new situations that arise today um because they formulate a theory of reality that makes predictions and that's interesting so these systems of thought and and they're more than that but they are systems of thought they are still relevant to us today they and they allow

us to make these predictions and if they are proven right as I think they will be it will let lend a hell of a lot more weight to idealism versus physicalism uh and that'll be interesting to see what happens in our culture thanks for watching sorry that was so long-winded and there's an even longer version because there's so much that one could say about this but uh hopefully this is the right length take care everyone see you next time foreign