As we shape AI, AI is reshaping us

Funny story. I just decided to ask ChatGPT what's it like to be a bat? Let me just read a couple things. It might be hilarious.

Planetary computation is the way that the world is evolving and that, like, it's kind of like a Gaia theory, I'd say framework.

Okay, cool. We made fire from rock. We made home. We have shelter. Great. Now we could like, oh, great. Draw caveman drawings. Oh, like, you meant this. Okay, like you drew bird. Oh, that's Bern. Like, we got to this point where, all right, we have all these different.

Languages and it's because, like, if you just look at things on a pathophysio physiological basis, you don't actually see the homeostatic patterns. You just see, like, physical structures of organs, and you don't see what the organ structure is. And the homeostatic function is what's evolving, not the organ structure.

We produced a digital version of an eye through like an evolutionary algorithm.

Our.

Are we talking about different languages here?

All right, everybody, welcome to the ABCs for building the Future podcast. Today I have myself, Robert, Jonathan, and Dean, my co founders on epistemic Me, our project to solve AI alignment and more. We bring on researchers, developers, scientists, founders, entrepreneurs, leaders who are on the front lines building future people with big stories, doing big things with big stories to tell. Today we're going to be nerding out about a talk by this professor, Benjamin Bratton. He had a talk which was called A Philosophy of Planetary Computation from Antikythera to Synthetic Intelligence. And in the first, like five minutes, he mentions epistemic or epistemology many times. And so this really intersects with what we do and kind of philosophy at the intersection of technology and AI. Really great talk. And we all listened to it as a team. And today we're just going to kind of riff and debrief some of our thinking. So the very first thing I guess I should say is this talk was about. Actually maybe we should just go around and do a round of like. What was the message you got from the talk, Dean? You were kind of talking about this before we started. Do you want to go first and then maybe Jonathan and then myself, and then we can go into some of the topics that I've listed out.

Yeah, I think there.

I heard.

I heard two things from the.

From.

From listening to the talk. One was the critique of epistemic institutions, which, I'm going to be honest, I don't think I really fully groked the argument there. And I'd be curious to understand what you guys have to say on that. And then the second, the second part of it was this idea of emergent intelligence, planetary level intelligence, which I thought was interesting. I didn't, I couldn't. I actually don't know too much about Bratton, this background. So how much of it was kind of this hypothesis on what is intelligent from a philosophical perspective or what's an interesting metaphor to apply to how we might view intelligence operating at the planetary scale? I wasn't really able to distinguish which side it fell. I thought it was a really useful way of framing. I've heard this idea from many other philosophers before. This idea of like hierarchical, hierarchical intelligence or intelligence emerging at different levels of abstraction. So all in all, yeah, it was a great listen. I actually just listened to it coming up from slow. But yeah, that would be my high level thoughts. And then we can deep dive on some of the specific topics.

Cool. So criticism of epistemic institutions and then kind of like, I think he, in the talk, he kind of talks about what you talk about as I think scaffolding and technology evolution, which we'll dive into a little bit. Jonathan, want to go next? Thanks, Dean.

Yeah, so let's see. I think he had this hypothesis that planetary computation is the way that the world is evolving and that it's kind of like a Gaia theory, I'd say framework where what we see happening with technology and the evolution of biology through time was like an early phase of the planet evolving. And then computation is kind of like taking it to this next level where everything's becoming interconnected. And he, he like kind of projects into the new sphere, which is a sci fi kind of like futuristic term for like when all beings on earth are, are intersubjectively connected through this computational kind of substrate. So and he says that the, the, this like computational evolution is kind of like unstoppable. It is like, you know, and, and the Anti Kythera name comes from of course, this Greek device that was found from about 200 B.C. that would track the planetary motions. So like in the sky. Right. So it was like a computational device that was like far beyond the capabilities we thought the Greeks could do at the time. And that's the name of the foundation and it's kind of like the metaphor for his theory. And he started off with a lot of like rich language around philosophy. I think he comes from a social sciences background versus a hard sciences background. And so he was like really kind of like putting some of the tools they use in social sciences as a frame for how to look at computation, how to look at the way things are evolving. And then his thesis was that this stuff is happening so fast and it is so complex that we need new concepts, new philosophy to be able to have explanatory power, you know, that can even tell us like where things are going. And that's where he introduces the needs for epistemic institutions, which is clearly kind of like a. Like a knock on the university and academia structure, working on the wrong problems. And that he wants to work on the right problems and have philosophy as a tool to give us language to work on the right problems.

The Antikythera thing was super cool. I think it's a cool name. It was hard for me to say at first, but yeah, the history of it is quite fascinating. And I guess for. Thanks, Jonathan. I guess for my own takeaways from it. The biggest. So, yeah, two takeaways. One was the criticism of epistemic institutions which I kind of mapped to. We live in capitalism. We are not able to sufficiently allow science to answer the questions that need to be answered within capitalism. That was my high level take. We could talk about that a little bit and discuss debate that. The other take was that really big thing that I got out of it was apes made fire from rock. Now we put intelligence into rocks and now we got to figure that out. And he has this entire idea of scaffolding and technology evolution, which poses, at least hearing him talk, this is what I was thinking, poses the question of like, hey, we keep centering ourselves as humanity of our own evolution. And we look at evolution and. But actually, aren't we. Maybe we're just a blip that helps materialize planetary computation the way that the planet and the universe needs to actuate. So instead of looking at humans as the center of everything, just seeing us as also parts of technology evolution or part of the evolution of the universe and playing a role there in things like intelligence, sense making, creating language to make sense of things and then artificializing intelligence, which we'll get into a bit. But those are some of my takes on it that I got out of it. He also talks a little bit about assembly theory. He also says some interesting things. Like one thing he said was sciences are born when philosophy learns to ask the right questions. Philosophies are born when technologies force the birth of new languages. And one thing he really said was we need new nouns to describe this. This does intersect a lot with discussions we have in Slack where we see that. I think, Jonathan, you mentioned this. Was it the CEO of DeepMind that mentioned that we really need philosophy to kind of. He didn't say this, but, hey, philosophy, could you level up your game so we know what we're doing on the technology side? I don't think he exactly said that, but something like that. What was that, Jonathan?

Yeah, pretty much. I think he said there's an urgent need for philosophy to solve the problems of AI and computation. And this is the guy that makes Alphafold. He needs ways to conceptualize how proteins fold and how cells work and stuff like that.

So lots of topics, so many different things. And I think maybe where we start is that criticism of epistemic institutions. Dean, you kind of mentioned, like, maybe you didn't quite grok, like, what the message was. Jonathan, do you want to start with the whole, like, how did you receive that. His criticism of epistemic institutions? And then we could layer in discussion from there.

Yeah, you know, it was a really interesting evening for me because I saw the event live at the Long now foundation. And this was like that week that deep, deep seat came out and everybody was talking about this and, like, the world's changing so fast and. And then, you know, so that was. That was the premise of the presentation. And I wasn't expecting it to be about AI, actually, but it was all about AI. And then at the end of the presentation, he introduced his organization, this anti Kythera, you know, organization where he has these kind of. What do you call them, studios in different cities. He's clearly trying to create a new kind of epistemic institution. And so I think he's got a certain thing in mind when he's going to bring together great brains that are working on problems that are related to how AI impacts human evolution. And he's going to create grants and sponsorships and get really smart people in these different cities to brainstorm and rethink the philosophy of what's happening and run different experiments. You have a lot of examples of the things he's already done in different cities. So I think what he was calling systemic institutions he presented pretty concretely as, like, what his organization is kind of trying to do. And I think he's very disillusioned with, like, the. I think maybe it was UC Santa Barbara, where he works.

San Diego.

UC San Diego. Okay, cool. Yeah. Where he just doesn't think like, you know, they're able to kind of reformulate the questions of the day and meet the challenge. Technology is giving and pushing on the questions of philosophy.

I guess the question is, like, what new epistemology or what new process is he proposing that would prevent his institution from repeating the, like, the exact same problems of the institutions he's criticizing? I think there's like, interesting examples of like, institutions which operated with completely different epistemologies. For example, like AT&T, Bell Labs. I think we could definitely go there if we want to like, compare and contrast all these different examples of epistemic institutions. But I'm also, I think that's the one thing I was struggling to understand. It's like, I mean, like, there's been many institutes like the Santa Fe Institute, which have just tried to get a bunch of smart people together. But ultimately oftentimes those under also fall under dogma or they fall under the same problems as like the university. So I'm curious if he talked about it all, how he plans to avoid that.

I don't think he talked about it in detail. I just reviewed the talk like a couple hours ago. It's a little fresh in my head. But his main thesis that I remember was he talked about how architecture, physical architecture, has benefited from this money flowing in and humans wanting to figure out, hey, how do we exist and organize ourselves with space and technology and being alive in respect to nature? And so he argues that architecture has been given space to breathe. Those are my own words. And that technology. And then also at the intersection of philosophy and I think knowledge, truth seeking as a process is a little capped in the software space in particular, he said, particularly with software. So he said software needs also a home like that. So that's his whole thing is, hey, Antikythera is going to be. And this is the big question that I don't know the details to of like, did he just find a bunch of rich people that were like, yeah, you need to go answer these questions and not be beholden to anybody. Here's a bunch of money. But there is some. So I don't know the details of that and how the double click, triple click, quadruple click details of like, how can we actually ensure the incentives are not misaligned so that you are actually finding truth? So I wouldn't be able to talk about that on this pod. But more so his thinking that current epistemic institutions like universities, academic research centers, and I've known this because my brother is a PhD in biophysics. So I've talked to him about this. I've talked to other friends who've gotten PhDs or masters. It becomes like a political game of let me get a grant to then be able to overlap with the question. I really want to answer, but then I kind of have to answer this other thing that has kind of capitalism interests. Right. So then I almost dilute my truth seeking process or my knowledge seeking process in this system. So then the question becomes like, well, at this point in time, with the birth of superintelligence, this, this is really not the right time to be capping our truth seeking. That was what I got out of it. So Jonathan, did you have anything to add to that or react to that?

Yeah, I think that's really well articulated. I was just looking at the language on his webpage. The speculative studio culture comes to mind. Foundational questions of what computational systems are and could be, examines what is necessary and what isn't, and maps scenarios to test chart trajectories accordingly. So I think that's kind of like the commitment he's made to his stakeholders who are seem to be like a large European sponsor for the project. It's really interesting and I think how you described this process of grant seeking. Everybody knows how broken it is and how disjointed is from the real questions you want to ask. I think that's the kind of thing he's trying to present a hypothesis for other ways to do that. But, but one of the key things that came across when I was like looking at the comments on the YouTube was that people with the, with that were using language of hard science were very critical of like his approach and the way he was. And I think he kind of exemplifies like this divide, that philosophical divide between the hard sciences and the social sciences who literally have to use different philosophical tools to solve their problem. And I think he's kind of saying like, hey, computational people have you hard science problems all the, you know, for all of history, like let's, let's bring some of these social science constructs such as philosophy into the, in the equation now because technology needs it. And that is a controversial take, at least from the academic perspective.

It's interesting that we would say that software or I guess technology is more epistemically bound, because I actually don't see that at all, at all. I feel like, I mean the recent advancements in AI are going all the way back to the transformer have come out of, you could argue slightly more capitalist institutions who moved at a faster pace than the university system. So you could almost maybe argue that like capitalism or the allocation of resources is proving to be a sound epistemic strategy for producing better and better technology. That's almost like the perspective I would put on it. As opposed to. Yeah, as opposed to needing. It's more of. Yeah, it's actually less so about the resources, let's say about the grant seeking and more so about the, the rapid rate of iteration and like being able to like show that you're correct. And that's one thing that capitalism seems to offer very well.

Yeah, it's interesting that you present it that way because I think it's been this kind of like there's been an intellectual battle about like how technology should evolve and analytic philosophy and capitalism has been what has worked and what has always worked. And then all along the way there's always like a philosophy call it, you know, from the direction of continental philosophy or system sciences or consciousness philosophy that says no, this is inhumane. We're like, we're not considering the human problem in this both in the capitalism and in the like the progression of analytic technology. But you know, like what produced transformers, you know. And I think that was clearly just like capitalistic investment. Bigger, bigger. Bigger analytic technology, faster hypotheses being tested like you said, and results.

Yeah, I think that's interesting and intersect I guess adding to the conversation on like the other take is I've been reading this book Super Agency. We have it in our little book club if anybody wants to join feel hit us, hit us up and they. Reid Hoffman actually argues a similar thing. Dean. And he argues that. How so? He argues that there's these. Let me just describe these nouns he uses. There's broad brush kind of labels of the different four types of people in the discourse of AI at the intersection of like ethics and rapid rate of technology evolution implications of humanity in the universe. So there are doomers who believe we're on a path to the future where worst case scenarios, super intelligent, completely autonomous AIs that are no longer well aligned with human values may decide to destroy us altogether. Except perhaps, and he literally says this except perhaps for a small contingent of tech bros whom they'll keep around to do the vacuuming as revenge on behalf of Roombas. I thought that was hilarious. Gloomers, it's the second broad brush label are both highly critical of AI and highly critical of doomers. And in their estimation, the doomer outlook serves dual purposes. First, it exists as a tacit endorsement of AI. It's so powerful it just might destroy us. Second, its long term and abstract nature misdirects attention towards the future where our real priority should be more near term. AI risks loss of potential jobs, disinformation, misinformation, amplification, of existing systemic biases. So that's glimmers. They favor a prohibitive top down, let's stop everything kind of approach or highly regulated approach in contrast. So there's a third one, Zoomers. Zoomers argue that the productivity gains and innovation AI will create will far exceed any negative impact it produces. So they're skeptical of the idea of precautionary regulation. And then finally there are the bloomers. So the super optimistic people, kind of like Zoomers, they're optimistic and they believe AI can accelerate human progress in countless domains. And at the same time, they recognize that a technology as transformative as AI cannot and should not be developed or deployed in a unilateral fashion. So Hoffman kind of sees himself as a bloomer that very optimistic and tying back to what we're talking about with how capitalism has really funded rapid iterative experiments, the evolution of Transformers. He has this term called iterative deployment, which is OpenAI's model of we're just going to ship and get as many users on earth to use it, and in that way we can have them participate in the creation of this new technology or influence it. So I thought that was a very interesting take, which is like, oh, actually, if you think about it, and Benjamin Bratton kind of talks about this too, it's like you can be a user for 20 bucks a month for a monthly subscription, and then you could actually, if you wanted, influence the roadmap. And in a way it's like, well, there's hundreds of millions of users all across the globe, very diverse perspectives, very diverse backgrounds, kind of funneling in feedback to evolve at least these products from that OpenAI use case. I thought that was interesting. I haven't quite formed an opinion on that. Dean, I see you nodding. I'd love to. What thoughts you might have.

I, I love that little tidbit about like you mentioned, like subscriptions as a way to democratize access to technology. And it's interesting because I, I, I work at Stripe and so I'm very familiar with the subscriptions business, the way that that works. And yeah, I mean, in many ways actually I feel like subscriptions gets a bad rap. Like many people consider it to be like a poor user experience or it's ruining the way in which I buy products. I can't pay a single fee anymore to access my product forever. I'm like kind of bound to this like continuous income cash flow. But yeah, it's actually was a very cheap way to democratize intelligence. You could argue that if I had to buy the product one off the first iterations of ChatGPT would have been prohibitively expensive. Now they can actually launch these different subscriptions and without knowing the cost economics of AI, in the future they have the capacity to launch new subscriptions, change the tiering model. We actually see that with Pro. And plus it just, it allows you to monetize and economize a technology in very different way, which, you know, oftentimes we don't see as being important, but it is very important. And in this way it is one of the fastest technologies to, to democratize by any means.

Yeah. Just to put a number to how fast I think it was they got after a month of ChatGPT launching, they had 100 million users, which is really fast. Right. So one could zoom out and argue what other technology has allowed such expansion of democratization of influence of user feedback towards a product than ChatGPT or AI. And also what's interesting, on the other hand, you always hear this news of like Sam Altman's evil, they're going to destroy the world, that kind of thing. Obviously with these new nouns. I think that's more of the doomer. Jonathan, I'd love to hear your take on this.

Yes. I think if you look at those categories or the deceleratorist or accelerationist point of view, I think it matters as much how you believe in the status quo and the forces of progression to this point. And I think if you look at capitalism and analytic philosophy that have produced a lot of technical innovations, one of the things that they, or two of the things they don't do well, one of them is that they don't model subjectivity well, they don't model, they just sort of don't have a metaphysical premise on the world. They assume that all observers observe the same world and the cause, effect relationships are the same for all of us. And two, it doesn't model evolution well or at all. Like there are literally no, no strong theories of intellect, philosophy of evolution or how we got here. And that's, that's teleological and, and in fact I think Bratton in the presentation like used artificial artificialization as his construct of teleology and teleological progression. So that he's bringing in philosophy that he thinks is like where he's saying the status quo is going to solve these problems. Right. But I think where it comes to that rapid feedback loop and like, okay, how do you really interpret what values the people are getting out of the software that they're using in terms of just product management and Value definition and constructing the language to get feedback. I think that's where, when you have such a mass democratization of tool and you're getting feedback from so many different positions and perspectives, how do you collate that? And that's what sort of epistemic me is about, is let's give a specific lens or frame for every single individual and their individuality and their subjectivity to contextualize the feedback that they're giving. And the hypothesis is that that's a more meaningful way to interpret the data and to make the next change and test the next set of hypotheses in system. That's how I'd look at it.

Yeah, that was super interesting what you said about how analytic philosophy doesn't address evolution. I wonder if we could double click into that and explain in your perspective why that is. Because actually I don't think I can put into my own words why analytic philosophy wouldn't be able to describe evolution.

Yeah, I think when it comes to fitness functions, and I will say in the presentation they mentioned assembly theory, which the whole goal of assembly theory was actually to put out a definition of evolution and the founders of it said we don't have one yet. Like, so this is, you know, this is a concrete mathematical definition of evolution and it's coming from people in chemistry, interestingly in the chemistry field. So, you know, so I think it's like a grounded perspective to say like there is no universal analytic theory of evolution and there just isn't one. And why. Right. I think you have to look at like, if you're actually going to contextualize like population and fitness, it really comes down to modeling consciousness and subjectivity of the fitness criteria that are actually selective of environment. And I think one thing that's illustrative of this is the ORP ontology. Ontology, Wait, ontogeny recapitulates phylogeny, which was this theory in the 19th century when they would cut open cadavers of animals, they would look for progressions like, of like from fish to mammal to see like do the organs inside of the animals represent the evolutionary changes that we'd expect? And they don't. And it's because like if you just look at things on a pathophysio physiological basis, you don't actually see the homeostatic patterns. You just see like physical structures of organs and you don't see what the organ structure is and the, the homeostatic function is what's evolving, not the organ structure. So they're kind of like, you know, just taking an object, a view of objectivity that doesn't actually lend itself to theories of evolution. And so I think it like, I think that's why there's no, there's no theory of evolution coming from analytic philosophy.

Would maybe it's helpful for the, for the audience, for us to define analytic philosophy. Like how would, how would you describe analytic philosophy?

Yeah, I think, I mean, you had Kant and then you had Boole responding to Kant. So Boolean philosophy is what we use in our computer systems today. Frege was the German philosopher who really kind of took epistemology into a mathematical domain. And then there was a kind of constant march of analytic philosophers between Frege and Quine. Quine wrote the two dogmas of the empiricism. And I would say our current modern philosophy of science hasn't evolved much past Quine. It he kind of like collapsed metaphysics. He says metaphysics doesn't matter, teleology doesn't matter. All that matters is epistemology and these reference frames that we can use to define cause, effect, relationships. And there would be people that say there is no philosophy other than intellectual.