Hod Lipson - Beyond Biology: AGI Minds in Competition (Worthy Successor, Episode 32)

**Daniel Faggella** (0:08)
This is Daniel Faggella, you're tuned in to The Trajectory, and this is the 32nd episode of our Worthy Successor series, where we think about the permutations of life that we'd hope to be able to bloom beyond humanity when we're long gone. We've had many different speakers with different backgrounds here, from a Michael Levin in biology and an Ed Boyden in neuroscience, to a Nick Bostrom who's more of a philosopher, or Yoshua Bengio, many, many an interesting speaker. Robotics hasn't been a main push for guests on this series, I suspect that'll change over time, but in this particular episode, we go right to a leading thinker in robotics who's been for decades pushing the limits on what robots can do, and in my opinion has some of the most sharp ideas around where robotics and AI are ultimately going to take us, and that is Hod Lipson.
Hod is a leading thinker in the domain of robotics and even artificial life. He is also the director for the Creative Machines Lab at Columbia University where he's taught for a great many years. In this particular episode, we go in with Hod and get his conception of where AI is taking us and what kinds of life he would hope to blossom beyond us. He brings up an interesting quote that will open the doors for us here, that if we expect to wake up AI and tell it to solve for cancer and solve for energy, but not really care for its own needs or solve any of its own unique problems, that that's probably an untenable position and that almost certainly the conjuring of what we are going to conjure will force us rather rapidly to co-evolve with it as there really won't be any other choice. A remarkable degree of frankness from someone who has an academic position. Not everybody is as frank as he is. I enjoy this episode a lot. I hope you will as well. I'm going to save my comments as usual to the end of this episode. So stick around for that. Without further ado, let's dive in. This is Hod Lipson here in The Trajectory.
So, Hod, welcome to the show.

**Hod Lipson** (2:01)
Pleasure to be here.

**Daniel Faggella** (2:02)
Yeah, there's a lot to unpack with your thought.
And before we dive into the worthy successor questions, I sort of want to unpack the way that you think about intelligence. I think what separated your work from a lot of other thinkers in my book is sort of a grounded grasp of what intelligence is as sort of an expanding process. And so I'd love to have you maybe open us up with, when you think about intelligence itself, how do you explain it conceptually? And then maybe we can get into some of the subcomponents here before diving into the bigger worthy successor questions.

**Hod Lipson** (2:36)
All right. Well, that's a very, very big question that you just asked. There's a lot going on in this umbrella of intelligence. But for me, there's a couple of interesting aspects. The first one I would say is the fact that intelligence is not just one thing. It's not that we're trying to build a ChatGPT or this or that.
It's a whole kingdom. It's a new kingdom of life that has different forms of intelligence, just like life on Earth right now has intelligence in lots of different forms. So that's one thing and we can talk about that. But the core issue around intelligence, I think, is that people think a lot about how AI views the world and world models, what it can do and cannot do and can it help humans? Can it do this task? And that tells to me the fundamental question to ask about intelligence is how does intelligence see itself? And that moment where the machine is not just thinking about the world and trying to do this task or otherwise beginning to think about itself, that's when it really becomes intelligent and sort of the sentient meaning of the word. And this is where a lot of magic begins to happen.

**Daniel Faggella** (3:52)
Well, yeah, we'll use this as our gateway, Drug Hod, to what you had just said there. I think, you know, you think a lot about phase transition. I think you've kind of, you know, you see phase transitions as important. And I think most people would say, and in fact, some people, the anchor of intelligence for most people is humans. I mean, it's hard to blame us because on some level, we're the most understandable intelligence that we can grasp because we at least kind of understand what homo sapiens want. Some of the time, better, better, at least we tell ourselves that whether it's right or not. But, you know, we don't live inside of, we can't snap our fingers and live inside of, you know, a blue whale or some kind of lizard with no eyeballs or some other intelligent entity at a strata of nature for which we have no sense perception to be able to access in the first place. So our imagination is bounded naturally by our form. And some people would say, well, the thing that differentiates humans, I mean, maybe Aristotle would have a similar idea, which would sort of be this reflexive capacity. Like we can think about our own thinking. We can make reasoned decisions, be held accountable. And so what you're articulating is sort of that core threshold as well. But I think you see the idea of self-modeling happening at various scales and potentially happening in the future at new scales. Walk us through this reflective idea as far as you see it in intelligence.