Looking At Max Tegmark's Vision of AGI 7 Years After Life 3.0 Ep.247

the questions about the dangers of the release of superintelligence now what happens when you put a Prometheus in their hands and what does that mean I'm

yes we want it in alignment with human values and his approach to that is part of a larger philosophical discussion about the nature of intelligence and

[Music] life hey what's going on everybody today is July 16 2024 and you are watching and listening the daily ey show live and

today we are going to have an awesome conversation about Max tegmark's vision of API uh AGI AI uh seven years after he wrote a

book life 3.0 and luckily today we have an expert in the room that that knows all about it but including with that expert who is Andy by the way uh we have Jun me and Beth and um Brian and uh we

we all know a little bit about it but this was Andy this was your idea to talk about this book I will be the first to say I did not know this name when you first said hey let's let's do a show about this I was like who and what and

please chat GPT and perplexity please help me um but I will tell you I i' I've learned a lot I'm actually really excited to have this conversation so I'm really excited for you to kind of bring this in case you know lots of people

from our audience are not familiar with this book from seven years ago and what it really means for our current you know world of AI and and what's going on so

with that I will have happily turn it over to you Andy to uh impart all the knowledge and wisdom well I won't be the one imparting but I'll try to collected

I think one of the virtues of the daily AI show is that we introduce topics that you know at at the outet we don't really know much about our even ourselves we're

learning alongside each other and along with the audience in many respects and I had the opportunity last year to read

this book uh life 3.0 by Max techark and it really was foundational to my understanding of AI and I was surprised

later to find out that it was written in 2017 but let me tell you just what it is it's it's a thought-provoking exploration of the future of AI it's

about the benefits and the risks the ethical considerations and the technical challenges of controlling a super intelligent AI so who's Max tegmark and

why is he in a position to comment on this even so he's uh he's from Sweden and he showed an early aptitude for technology he sold a word processor and

a 3D Tetris light game while still in high school so one of those guys like Mr reindeer he pursued education in Sweden he got a master's in engineering physics

from the Royal Institute of Technology and and this is important a ba in economics from the Stockholm School of economics so he's very Broad in his

academic Pursuits and and his study of Economics I think is relevant to his posture as someone who can comment on the future of not only the development

of AI but the future of life and and how things work in our economies he got a PhD in physics in 1994 from UC Berkeley

and then he joined MIT as a professor in the department of physics he's been elected a fellow of the American physical Society uh awarded the Royal Swedish Academy of engineering's gold

medal in 2019 and in 2023 he was included among 100 most influential people in Time Magazine um uh and and specifically with

respect to AI okay his his research at MIT is focused on cosmology which I think the nature of reality and the Universe on physics and artificial

intelligence so want to paint the picture of a kind of a Buckaroo Bonsai superhero guy uh you know who's really studied a wide range of subjects and is

a major com commenter and influencer in the world of artificial general intelligence and his approach to that is part of a larger philosophical

discussion about the nature of intelligence and life and um recently you know uh open AI kind of laid out a framework for our approach to artificial

general intelligence and what their approach their framework as they've described is is oriented really towards the Practical perspective on the

evolution of the development side and practical benchmarking of how we make those stages of progression towards that whereas Max tegmark thinks about it in

this broader philosophical and futuristic perspective on the overall evolution of intelligence

so he classifies life and and intelligent life particularly in three different phases and that's where life

3.0 comes from so life 1.0 is biological life think worms and cells and so on that um that evolved its Hardware it it

has no control over its hardware and its software also was evolved through natural selection and software would be sort of the nervous system and and you

know how it thinks and so on ultimately uh you know not the worms they're not thinking but they have a nervous system um and then Life 2.0 is

where we're living right now and that's a cultural influence on life that allows through education and through the fact

that humans are the only ones that advanced their intelligence in life to this stage where humans actually pursue

pedagogy we teach our Offspring and we learn ourselves and we change our software through that learning process and so in effect what he's saying is

that we can't change our Hardware yet but we which is our body right we can't change our Hardware we can add technology on the outside but we can

redesign our software we can change our minds so that's life 2.0 then life 3.0 is a technological kind of life that can

design both its hardware and its software and that's what Advanced AI is going to allow us to do um

so this concept of 3.0 is relevant to artificial general intelligence and and in this book and it has eight chapters I think uh and and we could recite those

but in this book he lays out um uh in chapter five uh the story of the Omega team which describes artificial general intelligence being released into

artificial super intelligence the distinction between those two I think is generally understood artificial general intelligence is okay a an a a technological

intelligence that can perform at about the level of a human so it has reasoning capabilities it's self-directed it can do things uh

and a super intelligence a SI is one that can recursively redesign itself both its hardware and its software so

that it advances in this iterative fashion at the speed of technology that is far faster than the speed of evolution all right so big ideas and and

I shared this on an earlier show this story the Omega team and I'll just give you an idea of of this uh progression this rapid progression once artificial

superintelligence is achieved so the the story that he puts in this chapter 5 is about the Omega team and it illustrates the idea of an intelligence explosion

once an AI system reaches the capability to design and improve itself so this AI that the the Omega team develops in is called Prometheus

and it advances from subhuman capability just superhuman capability within the first day of its release by recursively

optimizing its own architecture in multiple Generations so here's the story quickly uh on the first day at 9:00 am Prometheus version 1.0 is

launched it was slightly worse than the human programmers at AI development and it was being asked to focus on it own development at this point so within the

first hour Prometheus delivered version 2.0 of itself which was slightly better than version 1.0 but still

subhuman but just four hours later at 2 pm version 5.0 was released and it had uh it had blown past the team's performance

benchmarks it could no longer even be evaluated by the human team that was overseeing it and it was Exel ating and by Nightfall Prometheus version 10.0 was

deploying and the recursive process was continuing and they were ready to deploy Prometheus in the pursuit of certain objectives that they had including

making money to support the overall objectives of the Omega te now I want to jump ahead to say that control of such a

you know an explosion was very important in this story and the story progresses and I encourage you read it but the story progresses it's not a dystopian

story it shows how AI with the correct controls was able to create a very beneficial environment in this context

of Life 3.0 for the humans who are you know on the planet at the same time as this super intelligence um and and uh so let me

just uh wrap up by saying uh one of the ways that they controlled that is they really kind of put an airlock around

Prometheus and it could only reach out into the real world with supervision from humans so they they didn't give it

unfettered access to the world and and that prevented this the dystopian scenario from emerging and I want to just share with you a little bit about

Max tegmark's approach to the questions about the dangers of the release of superintelligence uh and and in 2014 he founded an

organization called the future of Life Institute and you can find that at futureof life.org and I want to just

read to you about the principles behind uh their approach to that this is now the um you know Max techark founded

future of Life Institute uh their approach to concerns about and regulation of artificial intelligence they also look at nuclear weapons and uh

you know bioweapons and so on but here I'll just read quickly and then I'll and I'll leave it to our discussion but our AI is racing forward companies are

increasingly creating general purpose AI systems that perform many tasks llms compose poetry create dinner recipes and write computer code some of these mods

already pose major risks such as erosion of democratic processes rampant bias and Mis information and an arms race in

autonomous weapons but there is potentially worse to come AI systems will only get more capable corporations are actively pursuing artificial general

intelligence which can perform as well or better than humans at a wide range of tasks this will bring unprecedented benefits but on the flip side more than half of AI experts believe there is a

chance that this technology could cause our Extinction this belief has nothing to do with evil robots or sent in

machines it is that AI can enable those seeking to do harm by easily executing complex processing tasks without

conscience and uh so militaries could lose control of high performing systems designed to do harm with devastating impact an advanced AI system tasked with

maximizing company profits could employ drastic unpredictable methods even an AI program to do something altruistic could pursue a destructive method to achieve

that goal we currently have no good way of knowing how AI systems will act when released because no one not even their creators understand how they

work boom mic drop uh that was awesome Andy as I fully expected it to be um one of the things you brought up and I just

wanted to bring bring up you didn't mention this part of the book at least from what I read again this is my chat gbt and perplexity summary of the book so perhaps it's wrong but my understanding is in the

book this Omega team when they deploy the super intelligence they do it to actually gain immense wealth so that they can then take that wealth and apply

it to whether it be Sciences human needs whether whatever that might be but I think it brings up the larger question it sounds like Max is also talking about

this in the future of life which is who who is in control and is that necessarily a great thing whether that's a corporation whether that's a country

you know is it in the best interest if let's say a a a the government the United States government ends up having they're the Omega team in in so many words right and they have a airlock

around a Prometheus like in the next several years let's say is is that good for all that you know the States might stand for and safe for democracy and helping is it is

it detrimental if you who gets to I guess what I'm getting is who gets to decide who's on the team because it I guess in the book they're they bring about the the brightest and they have a small team

and they do the right things I guess in the book like you said it's not a dystopian story but in real life it seems to me that it's a race

towards getting that airlock a around the Prometheus and it's not going to lead to more people wanting to share openly it's going to lead to what I think we're

seeing a little bit even with open the eye at others which is more of a closed off wall Garden where they're saying well we don't we no longer want to share what we're finding because it's at such

a level that it could be dangerous in our view if somebody else was to get and now there's no there's no sharing across the you know across the water or whatever the case is so I don't know if you want the touch on matter or anybody

else did but that was something I brought up in I just saw the notes that I thought was interesting in the book that that's how they chose to solve problems was basically gain wealth use

wealth to then go fix bigger problems yeah so I'll just make a quick comment about that that that's just one chapter in the book is the story which is a and

and a a large part of it that is really profoundly interesting is his Deep dive into the logic of how autonomous Weapons Systems

evolve in in practice in the hands of you know uh Bad actors against good actors if you can characterized that way and that how how

inherently difficult to Impossible it is to control that and what the what the downstream effects of that are so it's

not only about one Central AGI it's about many super agis out there many of them in the hands of people who have nefarious

objectives right right and one of the things that I think one of the things that like jumped out at me as I was prepping for the show um W is a phrase

that we see a lot that we want to make sure that AI AGI ASI is in alignment with human

values but we're not in alignment with human values right that's not an agreed upon thing across Humanity

which just calls into question for me like Yay let's do that but uh let you know are we going to have a conversation about what we think human values are and

and depending on how it's phrased right it it starts to say oh well that's not the way I phrase the human value that must be biased so now I'm I mean it just

like it is a can of worms and at the same time yes we want it in alignment with human values as opposed to not being align with human values right I

mean if the if it's door one or two we definitely want door one but uh I I think there's not agreement about what what's behind door one yeah for the for

the good of for the good of everyone is a relative term who for the who's who's getting to decide for the good of everyone you know and we see that in micro environments obviously all we have

to do is look back through history and all the atrocities and things where you know it's the Spanish Inquisition and and what the Catholic church was involved in through the years and stuff

like that just to see what for the good of humanity looks like misguided but at the time perhaps looks like this is the

best for everybody that's in history now what happens when you put a Prometheus in their hands and what does that mean for the seven billion humans who are

currently living on the US and continuing to grow in population so it's it's interesting I mean it's a really really like you said Andy I think you kind of said it it's a thought-provoking

story and I guess that's the point right it's the it's it it was to generate conversations like this and and hopefully thousands or millions of these micro conversations of people talking

about this openly and saying hey this is a complex problem this cannot be solved easily um but the best thing we can do is talk about it the worst thing we can do is go silent um because that that

won't get us anywhere that won't that won't get any hands Reaching Across any political or otherwise just using that as an analogy Andy made this point when he was

when he was giving the intro but again this was written in 2017 so he wasn't I mean it was published in 2017 so it was written whenever it was written slightly

before 2017 but also um like he he's clearly in many rooms of watching when it's happening but the conversations were not happening then hey it's

probably two years maybe 3 years maybe five at the outside and we're at this threshold that was not a predicted um like a as much of an agreed

upon prediction as it seems to be now and uh and I wonder how this phrasing or uh the way that the book is laid out I wonder what would be changed if it were

going to happen today if we're going to be written today yeah with the knowledge of of the world reaction to like 3.5 in from

November 2022 on so to speak uh would be interesting cuz it's not as if GPT 3 or two or one never existed of course they did they existed it didn't just it just

didn't happen in November 2022 these things have all been happening machine learning has been happening for decades but it was just this tipping point at which it captured the world's attention

and they went from like zero to whatever it was jimm me like what 1 million 1 million uh five days one million five days one million users I was like subscribers what's the word I'm looking

for yeah one million users in five days and if you look back at Facebook and other benchmarks earlier and add it just it wiped the floor with them hey before we get deeper into this conversation I want to remind everybody that if you are

enjoying this conversation and you maybe it's your first time to the show maybe you're a longtime listener like a lot of our folks in the comments who we greatly appreciate they show up daily and they add to the conversation we we really appreciate that you can head over to the

Daily eyes show.com that's a good landing page because what it'll do is number one it'll tell you how to see past shows obviously you can find them on YouTube you can find them on uh LinkedIn you can find them on Spotify

and other places but the daily ey show is really sort of our One-Stop shop also what you could do there is if you like these kind of conversations if you like the types of things that we're talking about and you're like hey I would I would actually take a little bit more of

this or maybe a different angle that's why we started our newsletter and so we put that out once a week on uh Sundays we are about episode or um it would be Edition number seven I think this week

um so we're just getting into it seven yeah betth is yes seven um and so you can go sign up for that it's obviously absolutely free we're only sending out that newsletter nothing else is going

out with it and then also you know if you are not in today's particular show but if you are working with your your company or you own a company you just work there and you're trying to figure out how to navigate you know AI for your

business then I would encourage you to reach out to us go to that website reach out to us as you can see on our screen here from all our little subtitles we all have different disciplines um that we come from I I focus in sales

obviously everybody else has different disciplines in Ai and what we ultimately try to do is bring all of our collective experience um from our team to help and work with other people's teams whatever

that might be so um go check that out want to make sure everybody knows about that and Bill please like And subscribe if you're enjoying this um we're here five days a week and yeah uh August 7th one will be actually our oneye

anniversary of the show so it's coming up soon um want to make sure everybody knows that Carl is also here as well he joined the a little bit a while ago but Andy was giving such an amazing uh intro there that we we didn't want to break

that but welcome Carl and I just ask you have you read did you happen to read this book did you know about it I was saying I didn't know about this before Andy brought it up or anything like that and what's been your you know your

perspective about it before kind of getting into Today's Show like leading up to the show I you know I honestly I I haven't read it or did I perplexity um

like summarize it so I just pretty much just listening to Andy um give his world famous descriptions of

it so um I think the the the interesting point here is that if we go back to the oh right um I can't I was trying to find

where I saw this number but it only said that 7.7 there are only 7.7 million chat GPT

plus users around the world and then it broke down other subscriptions just for open AI so if you think about that that's not very many people that

actually you know you I would say people who pay for it probably use it regularly just you just can't use it on a on that

basis on like with the with the limits so I would say though to get the actual buildup to to what Max was

talking about we need more and more people to actually use the tool or use the tools and if you go back to the router study

which what was it like 54% new chat gbt and then everybody else was like in the single digits or very

low double digits like 11 12% so uh that number to me was big because then it's like okay well it's really really really

early and I get to that Tipping Point you need way more users than that actual you like even Enterprise users I would say kind of fall in that bucket because

the Enterprises are trying to tell their people to use it they're using it kind of in very siloed ways and if you look at the

Enterprise um uh Revenue it wasn't it wasn't in the billions it was like overall it's a billion 3.4 billion but Enterprise Revenue was still pretty all

so you've got to think there is big big big big groups swaths of people that still haven't used it on a reg used an AI on a regular basis so I think that's

that's something important to to keep in mind too as we move as we think about this and what Max was talking about I'm

sorry as regards what Max is talking about um the vast unwashed population do

not understand how AI works or what its capabilities are will only be confronted by ai's outputs in the context of the

things that Max is afraid of which is okay in the hands of people who are you know trying to influence the political environment is misinformation on a grand scale they don't know that that's

actually happening you know how there's bot farms in Russia that are you know pretending to be humans well they don't need to have a a big building filled

with you know people sitting at terminals anymore if they have super AGI to do that on a massive scale globally in in order to influence opinions and

basically ship people's attitudes away from focus on the the sort of the terroristic aspects of what of of what

Russia's doing to Ukraine and instead blame it all on NATO for even pretending that it was okay to bring in Ukraine

into a defensive Alliance you know that you know all of that can shift Global attitudes on a scale unprecedent right and it's one of the

reasons that when you hear us talk about those things we're more often talking about China rather than Russia because China's game in this context is much

more elevated than we understand Russia to be right in terms of uh their ability to do something like this they're really where I believe

human people like there's not human people but let's just go with it there were human people in bod farms in Russia well no but you you've seen

you've seen okay we've all seen the the the the mobile Farms right so it's not the bot FS aren't going to be it's going to be the exact same thing where you

have like walls upon walls of mobile devices right the whole job of that is to you know whatever comments clicks whatever you need to get the engagement

level you have that you can essentially buy that and rent that so yeah for for it to have bot farms and so on it's yeah I but it's hard though because you have

to this is where the education component comes in where when you talk about the unwashed masses you you can't literally you

cannot believe anything you see or hear on on social like you can't you have to like triple check that you can't just run with it right because like the

there's so many things that came out like a new story would come out and then five different takes 20 different takes would come out of it and then the takes

is it it isn't a take some people say it's true but you just don't know right you you you like when you have to double triple quadruple check that um until it

actually you know it's like oh okay that's actually legit but by that time a lot of people have already bought into it so it's kind of the the dangerous point in with AI you can do it so so

quickly right also just want a quick uh push back that uh the unwashed masses are likely uh clearer about manipulation

than the washed MH masses I can't even talk but uh you can you can be part of what we're talking about as the unwashed mask even if you wash several times a

day cuz it's not about washing you know I just had a recent um you know what what was it two days ago like I said I'm overseas and so at some

point in the morning I on social media I caught the story of the assassination attempt on former president Trump in United States and you know Carl to your

point I guess like the first thing I did was go somewhere else to validate that what I saw was in fact true because the first because the first place I saw was on

social media and in this this charge climate you know political climate that we have in United States right now it's reasonable to assume that something like that could be fake there was nothing

about the video that would make me think somebody couldn't with with the right tools and the right you know the right desire um to fake something like that which is you know I I don't know both

sad and whatever but like that's the wor we live in and you know that the first thing I wanted to do was say wait hold on where is that well let me see that and so I'm constantly saying this to my daughter as well which is like she'll

say oh it's a you know she'll joke and say it's like a new trend on Tik Tok and I'll be like don't believe everything you see on Tik Tok you know I want to make sure my daughter's growing up in a

in a world where she has a whole different level of criteria that she has to live with in her life that I didn't have to live with at 13 you know there might have been a doctored photo or

something like that and you know whatever the case is um but nothing like what she is being um shown daily and and and it would be silly to think oh I'll

I'll put a bubble around her and I'll protect her I can't I can't protect her all I can keep telling her is just remember what you see might be fake it could be fake as real as it looks it

might be fake and she gives me a lot of that yeah yeah yeah Dad I know you think everything is fake I said no no no these are the conversations that I think parents are having all over the world is that you have to be really really

careful and it does kind of I think it to Circle back to the book obviously I think it does go in line with what Max uh seemed to be saying was there's a

there's a lot to to take in and it seems like he does a really good job of talking about the whole and not just talking about a particular piece which I

think is really really cool I think Andy you you sort of laid that out for us it I think it really is indicative of his background and his economics background and his MIT background and you know all

the things that he was able to do or has been able to do in his life um gives him this sort of worldview we were sort of talking about that yesterday in the show with this idea of generalist versus

specialist and it seems to me that he has enough worldly knowledge to be a you know um welcome the voice when it comes

to the future of AI you know and we need more people like him maybe not necessarily at his education but people who are the brick layers are the people who aren't necessarily being

touched by AI today to be part of this conversation because the last thing we want to do is exclude them by saying well it doesn't touch you today yes it like you're a brick layer don't worry

you've got probably a good 5 10 15 years before the robots can do your job better than you and can put up a building and they go I'm good I don't really need to know about the AI stuff it doesn't affect me at all in my life really like

how do we get more of those people into the conversation and get them to care which is tough because you know we all individually kind of care about our own lives and how it affects us personally so how do you get them to care about the

larger thing I think Indie in the book doesn't he talk about the next 10,000 years the next billion years and how we need to have a long-term view that perhaps sometimes sacrifices a little

bit of the short-term goal we need to be thinking way way down the road to figure out how something like a Prometheus is going to affect Humanity inter worldly

Humanity for for the odds you know yeah and you know humans are you know notoriously bad at long-term thinking so yeah yeah so it's easy to say that it's

really difficult to put into practice before we close out I just want to mention some other aspects of the book which are really kind of intellectually interesting so chapter six is on the the

future of life and then he discusses economic inequality job displacement and so on and has actual practical advice to parents all the way back in 2017 about

what the criteria are for the kinds of things that you ought to have your children thinking about in terms of the the you know decisions that you make about what you're going to do with your

life uh in the context of what's going to happen with the Advent of AGI uh chapter seven fascinating is about Consciousness

and he makes a very cogent argument about cons ious being substrate independent meaning it Consciousness

exists in our substrate which is neurons in our brain but that's not what's necessary and he and he kind of walks through I think a logical proof of why

Consciousness is not dependent on what substrate you're applying the Intelligence on meaning it can arise Consciousness can arise in nonbiological

entities so that's his life three it's a technological intelligence that doesn't necessarily mean that it has to include humans right intelligence can proceed go

ahead that I know you're wrapping up but I want to ask what's the difference between Consciousness and sensient are they synonyms well no um sentience it

means that you can sense something and I think it's arguable that you know sensors connected to that you know and and Consciousness is the is the

awareness of the sensory inputs right so they're they're related but not syon thank you um I do want to bring in a comment

from earlier let me go back and Pull It in and Jed it said we didn't know it at the time but 2017 was revolutionary for AI with the introduction of Transformers and Max tarks AGI insights yeah it's

it's interesting uh Justin I always think to I was working in business intelligence in 2017 sort of transferring back out of Entrepreneurship at the time and just starting to get into that data analytics

and machine learning like every every company we talked to was like we want machine learning in the same way people would say we're building AI today in 2017 it was it was ml ml ml everybody

wanted it they didn't know what it meant they just know they needed it for their business to get ahead and so we dealt with that a lot obviously so it's really interesting to think back to 2017 and

and to sort of watch this AI Evolution slowly roll out just in my perception but then your point to know that like what will probably be earmarked as

extremely important years the invention of the Transformer and how that basically was the birth of a lot of the AI that we know today and what they're capable of from the general intelligence so just want to highlight that comment

um really appreciate we appreciate everybody Jen and Justin and Cisco and and everybody else that's in our comment section you guys are the best um somebody was just saying um they'll have to go back uh to the beginning uh

because I missed the beginning of the show yeah absolutely we were talking about that earlier um definitely go back and check this one out and we've talked about a lot of these topics on other shows as well um coming up tomorrow we

have the news so every Wednesday we talk about the news um there's a lot of stuff going on in the news right now we talked a little bit about um strawbery we mentioned it as well as um the five

levels of AGI thata talked about um so those will come up I'm sure tomorrow as well as perhaps in individual shows because I think they probably take they need that attention at this point um and then uh Thursday we're going to be

talking about power power power more power because Ai and what any of this that we're talking about now for any of this to actually exist in the future um is going to take immense amounts of power what does that really mean where

is that going to come from you know I was saying yesterday quickly we are looking at power consumption on orders of not cities but countries so where does that come from it's a you know

who's going to do it how what types of power we're g to talk all about about that all about that tomorrow and then finally on Friday that's our two-e recap show so every two weeks on Friday we stop we don't have a new topic we

actually look back two weeks talk about all these types of topics new discussions that have come up news that has come up around these topics and just have a general open conversation so go check out uh go come back all week and

then definitely go check out the daily show.com um like And subscribe if you don't want to miss these shows like I said we do this five days a week um and we haven't missed a Monday through Friday yet we have gone through somebody

has been here every single holiday every single day and uh I don't know I feel like we want to keep that that street going so keep on coming back and hanging out with us guys um thank you Andy for

such an amazing intro and for bringing this book to our our attention and now bringing it out to the masses so everybody else can learn about it too so hopefully everybody will go check out the book and go uh go read it thanks

everybody uh we will see you guys uh tomorrow for the news until then have a great day [Music]

oh [Music]