Creative Machines and the Future of Robotics

Episode Highlights:

04:48 – Defining self-awareness as the ability to imagine yourself in the future

11:55 – The drive to build machines that can design

23:24 – Using robotics to investigate life, self-reproduction, and the possibilities of engineered systems

30:37 – Why passion matters when choosing a project or research direction

38:01 – Hod’s vision for accelerated innovation and engineers’ growing responsibility to guide AI

51:25 – The challenges of physical robotics and the importance of human movement data for training

Transcript

Hod Lipson: We have robots in our lab that, let's say, learn to walk, but we're not interested in how they walk in particular. We're actually interested in how they see themselves as they learn how to walk. So we can open their brain and you can actually look inside the brain of a robot, so it's actually transparent.

You can look inside and you can see how does a robot see itself? How does it imagine its own future?

Mike Massimino: I'm Mike Massimino, former NASA astronaut and Columbia Engineering professor. And this is Moonshot. NASA rejected me, not once, three times.

Three times I applied, three times I was told no, including being medically disqualified for my eyesight. I was and still am afraid of heights. But here's something I learned. One in a million shot is not zero. The odds are against you, do it anyway. 

Today I want you to meet Hod Lipson. Columbia Engineering professor, AI pioneer, and roboticist, whose creative, unconventional mindset didn't stop him from working on a problem decades ago. Robots that could think. He has built robots that can learn how their own bodies work, adapt when they're damaged, and explore where the machines can be creative, curious, and self-aware.

Hod, thanks very much for joining us. I know you're a very busy guy running our department and doing all kinds of things. So we want to talk about conscious robotics.

So back when you were an assistant professor, colleagues warned you, apparently, not to pursue conscious robotics. Is this true? Absolutely.

They said it was too fluffy? 

Hod Lipson: They said more than that. They said I'm not going to get tenure.

Mike Massimino: They said you weren't going to get tenure. But you did it anyway. So let's talk from the beginning.

What are we talking about here? What exactly is conscious robotics?

Hod Lipson: Okay, that's exactly the problem. Nobody knows what that means. And the whole issue of consciousness, emotions, sentience, self-awareness, the whole sort of thing is ill-defined, fluffy.

We kind of experience it every day. We know that we're self-aware, but it's very hard to define. I've been fascinated with this topic for we talk about why, but I've read all the books on consciousness, self-awareness, philosophy, theology, cognitive science, psychology, trying to find a definition.

What is, you know, we all experience it. It's important. What is it?

And I want to build it and couldn't find a definition. So there's all kinds of circular, fluffy definitions for it. So when I started working on it, people told me, look, it's a very fluffy thing.

It's for psychologists. It's not for engineers. You stick to hard robotics.

You stick to robots that do backflips. You stick to the real stuff. And let the other people deal with fluffy things.

Because even if you did, you weren't able to do it. Nobody would, you wouldn't know it. I mean, it's so ill-defined.

So that was the advice I got when I started my career. But I did it anyway.

Mike Massimino: Okay. So what kept you pursuing this when people told you it was a dead end? Because these people were trying to advise you in your career.

Hod Lipson: So first of all, I have to confess. I didn't only do it. So I did that and I did another 50 things, hoping that something's going to stick.

And this is my career advice, actually, for academics. Traditionally, people tell you, focus and do one thing and do it well. I'm saying, especially in today where everything is changing very fast, it's a huge gamble to put all your eggs in one basket.

And you actually got to do many things. Some things will work. Some things won't.

So I did a couple of different things. I worked on manufacturing 3D printing, creativity, which we can talk about, sort of machine creativity. But self-awareness sort of was the big holy grail.

I did that also on the side. And it panned out at this point now. Everybody in robotics is talking about it.

Mike Massimino: So there's a couple of messages there. I think one is don't necessarily focus in on one thing, but keep a lot of irons in the fire.

Hod Lipson: Exactly.

Mike Massimino: And that's not always the advice you get. Because there was like, you got to have what are known for, what is your contribution. And then, yep, so do you have a definition for conscious robotics after all this?

Hod Lipson: So I do. So here's my unromantic mechanical engineering definition of self-awareness.

Mike Massimino: I bet it's going to be a little romantic, but go ahead.

Hod Lipson: Okay. It's the ability to imagine yourself in the future.

Mike Massimino: Oooh.

Hod Lipson: So the ability to imagine yourself in the future. And the further you can imagine yourself in the future, the more self-aware you are.

So it's a difference between living in the moment, just being reactive, like a machine versus being strategic.

Mike Massimino: Yeah. Seeing the plan lay out in front of you, the future lay out where you want to be.

Hod Lipson: Where you want to be, what you want to do, what's going to happen to you, all the everything. By the way, when I say self, it could be you, your family, your organization, whatever you call ‘the self’, wherever you put that boundary–can you see yourself in the future? And to me, it's also useful because it's not a black and white thing, because the further you can see, the more self-aware you are.

And that's a difference between a child and an adult, or between a human and a dog. A dog can probably see itself a minute into the future, can taste food before it eats it, but it's not worried about retirement. We humans, we save for retirement.

So we can see very far into the future. I would say that's the main advantage we humans have over other animals. We're not faster, stronger. It's that we can see ahead.

Mike Massimino: Oh. And so you’ve been able to do that.

Hod Lipson: Yeah. So, so we've been able to, so it turns out you can go to an AI to a robot, any robot. We have robots in our lab that let's say learn to walk, but we're not interested in how they walk in particular. We're actually interested in how they see themselves as they learn how to walk. So we can open their brain and you can actually look inside the brain of a robot. So it's actually transparent. You can look inside and you can see, how does a robot see itself? How does it imagine its own future? And it's fascinating.

Mike Massimino: Yeah. This, you're making me think, because I'm thinking like when I was, when I was a grad student trying to picture where I would go after, you know, there was, there was a couple different directions I thought I could go into, into academics or into industry. But then in, you know, the astronaut thing is where I really wanted to go and then trying to imagine the steps or the, not necessarily precisely everything, but like what were the broad, broad strokes that would get me.

Hod Lipson: Exactly. So, so, so I think it's, it's, it is somewhat romantic.

Mike Massimino: It's very, I think it's very artistic.

Hod Lipson: But, but, but it's, but, but it's measurable. You can actually open up a robot and say this robot can see itself two and a half seconds into the future, whereas you see yourself three decades into the future. Right.

So, so, so, so once we have a metric as an engineer, we can start building things. We can start optimizing. We start measuring progress.

And that's what's happening. And so that was, that was sort of the way I set out to define self awareness. I have to tell you in the beginning, we, we made our first robot.

We could measure where we show that it was tiny bit self-aware. It could see itself a few seconds. What, what, one of the reasons why we're self-awareness is why it's, it's useful for my evolutionary point of view, the advantage of seeing yourself in the future as you become more resilient.

You can adapt. You can maneuver. You can dodge things. You can plan.

Mike Massimino: So you, now in your career, you're well established. I don't know if it feels like that, but I think you're, we'd say to say you are, right? You're, you're.

Go tenure. You get, you're tenure, you, you know, you've done really well in this area. Was there a time going back in the past, though, where you were a little afraid that it might go a different direction?

You sound very confident about now, of course, but back then, you know, when you're coming up with these ideas and these thoughts, were you, was there a bit of worry?

Hod Lipson: Yeah. There, there, there is a, you know, there's a, this is at the same time I worked on other topics like 3D printing, manufacturing, creativity is another one that was sort of very controversial idea that, that machine can, can be creative. That was also a taboo now.

People take it for granted. That's possible. But I think I was never really worried about it just because I had enough irons.

Mike Massimino: Mm hmm.

Hod Lipson: Yeah. That, you know, something, I had some solid conventional, you know, equations, you know, the usual stuff. But, but, but the real thing that, that, that kept me growing, the exciting stuff is trying to make a machine that is like a human.

Breathing life into matter.

Mike Massimino: Yeah. And then even though people at the time might have thought it sounded a little crazy.

Hod Lipson: Yeah.

Mike Massimino: That's where you have really significant contributions.

Hod Lipson: But you know, but you know, students did buy into this. Ah. So this is interesting.

So the establishment, if you like, said, okay, this is fluffy. This is not a good way to build your career. But students got very excited about this.

So I think, I think you get different sort of feedback from different generations.

Mike Massimino: Sort of from students who haven't been polluted yet. 

Hod Lipson: Yes. Exactly. 

Mike Massimino: And how to think.

Mike Massimino: That's cool. All right. So more on these ideas.

So you spent your career, we mentioned a couple of these ideas, but they almost sounded like science fiction at the time. Even when you came to Columbia, I mean, self-aware robots you talked about. But I remember you talked about 3D printing of food. Right. And I'm like, I don't know what kind of lasagna that's going to make.

I mean, you know, that machines that can design other machines. And you mentioned people thought that a lot of this stuff was crazy. Is there anything …what has surprised you the most?

But what has actually become possible? Is there anything out of these science fiction crazy ideas? Or maybe they were considered crazy. Is there one that surprised you the most?

Hod Lipson: Well, first of all, first of all, I have to say. Food printing. Surprised me in the sense that it did not work. It's still.

Mike Massimino: Ah. It is not easy to recreate.

Hod Lipson: Something about. So I thought food printing and laser cooking, which is another thing we've made. Laser cooking.

Laser cooking. Have you ever eaten a steak that's laser cooked?

Mike Massimino: No, I don't think so.

Hod Lipson: You got to try that.

Mike Massimino: They did. You have?

Hod Lipson: Yeah.

Mike Massimino: You just hold the laser to it?

Hod Lipson: No, it's a gantry. And it just cooks it. Blue laser cooks it inside.

Don't get me started.

Mike Massimino: OK.

Hod Lipson: But I thought this would be amazing. There'll be people buy this left and right.

Mike Massimino: Yeah.

Hod Lipson: And nobody wants to talk about food printing and laser cooking. This is to me one of those irons that didn't work out.

Mike Massimino: OK.

Hod Lipson: Because I spent a lot of time on food printing. And I thought that was going to be a big hit. So that's an example of something that didn't work. I still hope it will.

Mike Massimino: Yeah. Maybe it’s just ahead of itself.

Hod Lipson: Maybe. But the thing that surprised me the most, I would say, is machine creativity. It's this idea that a machine can be creative. And that's another one of those things. When I started my career as an engineer and I spent a lot of time designing things. And I got bored.

Because most of engineering design isn't exciting. It's actually you’re designing brackets, handles, all kinds of things. And I thought surely there's a better way to do this. Surely, instead of designing all these different things, how about if I design a machine that can design those things for me? And almost like you go one step up.

Mike Massimino: Yeah.

Hod Lipson: And once you can automate the design process, two things happen. One, you don't have to design all these mundane things. And the second thing that happens is that maybe you can design things better than you because I don't think we engineers, to be honest, we're slow. We get bored. If we can automate this, maybe a machine can design things that are better.

We get locked into conventional ways of designing things. Maybe there's alternatives. So I spent a lot of time trying to think about ‘what is creativity?’

I read all the books on creativity and design. And I didn't find any answers. So I got my answer for you for what creativity is also. A non-romantic definition. Because if you look at all the books on creativity, they'll talk about the muse and all kinds of philosophical things. But I think that creativity is really just the ability to fill in the blank.

The larger the blank, the greater the creativity. So it goes all the way from just filling in a word in a sentence, which is trivial, all the way to filling in the blank page. Which requires a lot of creativity.

And it turns out that's exactly how we ended up training generative AI. Generative AI is trained by filling in the blank and you stretch the blank and you train the AI to be more and more creative. And that's exactly how it ended up working.

Mike Massimino: Do you have an example of this?

Hod Lipson: Yeah. Well, the most trivial example is if I give you, I take a poem and I remove a couple of words and I say fill in the blank. And it doesn't require a lot of creativity.

But then if I remove a sentence out of a poem and I say fill in the missing sentence, it requires you to think a little bit more. If I give you just the prompt, just the first sentence and I say fill in the rest of the poem, that requires a lot of effort. But you can train that way. You can train a machine to fill in the blank. 

So that's for poems, but for images, it's the same thing. You take an image, you remove a pixel and say fill in the missing pixel. That's easy. Now you remove a patch, fill in the missing patch. And then you fill in a bigger missing patch.

At the end, you say fill in the blank page. And this is how we train AI to be creative with words, AI to be creative with images. And what people have not done yet, and there's something we're working on, is here's a machine, here's a design. I'm going to remove a piece. You fill in the missing piece. Not so hard. That's a, you know, apprenticeship challenge. Now I'm going to remove a bigger piece. I'm going to remove a whole mechanism.

You fill in the blank until I give you a blank page and I say design the machine from scratch. It's the same thing. You just learn to fill in the blank.

And with that, again, definition of creativity, you can actually train a computer to do it.

Mike Massimino: Are we close to this site? Give me a design for this and AI comes up.

Hod Lipson: We're not there yet. We're not there yet, but I think it's going to happen in the next two or three years.

Mike Massimino: It almost sounds like you're training it. Like you would train a student to think and come up with a design that's saying, this is a known design.

Hod Lipson: Yeah, exactly.

Mike Massimino: Come up with this part of it. It's just what you're going to do.

Hod Lipson: Exactly.

Mike Massimino: Exactly. Do you do this in your classes? The rumors that I hear from students.And now I'm going to use AI in my classroom, this semester. But some of this, oh, Hod Lipson, let’s I use it. So do you, do you, some of you do this as well?

Hod Lipson: I absolutely encourage students to use. I don't only encourage them, I give them a test they cannot do unless they're using AI.

Okay. So if you're not using AI, you're not going to be able to complete this assignment by next week. You have to, you need to sketch out 20 designs for potential robots that you're going to make.

Mike Massimino: In one week, let’s say?

Hod Lipson: Yeah. So if you're going to do this manually, forget it. Okay. So, so, and I want them to be crazy.

So you've got to work with the prompts and you've got to generate all these ideas. And then you've got to pick one of them and build them. But you've got to use AI to ideate.

Now, that's just sketches. What we haven't done yet is an AI that actually creates the three-dimensional designs of the nuts and bolts and batteries and wires. That's the details.

AI isn't there yet. But I think in the next couple of years, it's going to be there in a couple of years. Two or three years. 

Mike Massimino: We talked about things that failed and, and worked and some surprised you. Uh, is there any, is it just kind of, I don't know, is it happenstance or luck? Or is there a way to tell a bad idea from a, from an idea that maybe, you also said that's maybe not ready yet.

Right. So maybe there's a bad idea. Maybe there's a really good idea.

Maybe there's an idea that's going to be good. Is there a way that you can kind of judge that? Or is that just?

Hod Lipson: I think I, if there is, I don't know one. Okay. There, there isn't a way, I mean, look, the VC is trying to do this. Yes. I mean, nobody, nobody knows how to do this.

Mike Massimino: Right.

Hod Lipson: I mean, you just have to, if you ask a VC, it's one in 20 is a good idea. So I have 40 projects in my lab. Two of them will work out. And this is the, this is the benefit of being faculty. I have to say is that you can have many projects running simultaneously. And then some of them are going to, to yield and, and produce in some, and most won't, but you always got something.

So you got to have that plurality of ideas and projects running. Again, I see some faculty making the mistake of focusing. And they try to pick in advance two or three projects that are really going to succeed. And look, you're going to talk to other people. Maybe they'll give you other perspectives. Maybe they know how to do this.

I can't. I just have to do a, you know, just to see, see what sticks.

Mike Massimino: Yeah. But it's interesting because people, I think sometimes think they should know. And the reality is you don't.

So by having those more than one iron in a fire, having these multiple, and you never know, you're going to get something out of it, right? You might not get with your sounds like you're going to get something.

Hod Lipson: And even look, when I take it, let's say a PhD student, they will work on three projects, not one project, just so that they have that diversity of… this risk mitigation, if you like.

Mike Massimino: When people are telling you that things aren't going to work, so I bet you probably got opinions even now, right, from people who said this isn't going to work, that it's impossible. Is there something that you remember or is there anything that keeps you, what keeps you going when you hear that? People are telling you it's not going to work.

Hod Lipson: Well, first of all, nothing gets me going more than somebody telling me it's impossible. Just the challenge. All right.

So this is, this is, maybe that's a personality thing, all right? So somebody telling me machines can never be creative. Or even working with 3D printing, I had manufacturing people at the time say, hey, this is, it's just, it's not serious.

It's not really going to work. It's a cool thing. But, you know, CNC and we're going to keep making things the old way.

But it's, that sector is still growing 25% a year in terms of manufacturing and even in metal. So, I mean, there's always people going to tell you something's not going to work, but you're going to take it as a challenge.

Mike Massimino: That is not to get discouraged. Oh, yeah. I've seen this before.

We're going to go exactly. I love it. Okay.

We want to talk about the future and I want to talk about the past a little bit. But before I forget, this definition you have on creativity. You got a book coming up on this? Have you written about this?

Hod Lipson: Yes. I'm writing a book on curious and creative machines.

Mike Massimino: Okay.

Hod Lipson: And that's, that's.

Mike Massimino: It's in the works. It's in the works. Like, no kidding going to happen.

Yeah. Sorry, look forward to, we look forward to seeing that one. All right.

So let's go back a little bit. No, we'll go, we'll go forward again. But when you were a kid, where did you grow up? You grew up in Israel?

Hod Lipson: I grew up in Israel in Haifa.

Mike Massimino: Yeah. So when you were, you were there, I mean, did you take things apart? Did you, were you a curious kid?

What was going on in your head when you were a little kid?

Hod Lipson: Okay. What was I, was I, so, you know, again, I, I, a lot of people like to say, oh, they took things apart to understand how they work.

Mike Massimino: Yeah. Yeah.

Hod Lipson: I wasn't like that at all.

Mike Massimino: You get yelled at if you do that in my house. Say what'd you take the toast apart for?

Hod Lipson: Exactly. But also I, I don't want to take things apart.

Mike Massimino: Yeah. Don't break it.

Hod Lipson: All right. So, so, but I, I just like love to build things from scratch.

Mike Massimino: Yeah.

Hod Lipson: And my father was a physicist, but he has a hobby as a carpenter and he had this workshop. Nice. Hang out there.

And I would build things out of LEGO. Like, yeah. Now back then LEGO was very expensive.

I would get one set a year.

Mike Massimino: Okay.

Hod Lipson: Now my kids, they get one set a week. I mean, this is like a, and they never take anything apart. They never use it because it's all, you know, just get a new set.

Mike Massimino: Yeah.

Hod Lipson: But I had this one precious set a year.

Mike Massimino: Yeah.

Hod Lipson: So I had to reuse and rebuild and every, so, so I kind of, I think that's scarcity. So it helped me learn how to build things. And that's still what launched me to sort of mechanical engineering.

Mike Massimino: And that was, it wasn't, I guess it wasn't one of these kits, like build a rocket thing.

Hod Lipson: No, it wasn't. It wasn't originally a kit.

Mike Massimino: But, you know, maybe you made your own thing. Yeah.

Hod Lipson: But we have one thing for a year.

Mike Massimino: Yeah.

Hod Lipson: You know, it doesn't last very long and start taking apart and doing the things.

Mike Massimino: So many people were inspired by playing with LEGO.

Hod Lipson: Yeah. So for me, that was, you know, the, that was, that's a good one. It was a great way to start getting familiar with, with, you know, getting your hands dirty.

Yeah. You know, 3D and the whole thing.

Mike Massimino: Yeah. And that was when you were younger, of course, I guess, like elementary school and so on. And then as things progressed, what, what drew you toward engineering maths.

I mean, I had it, things progressed.

Hod Lipson: So when you were growing up, I remember always trying to, so I think I'm a descendant probably from an alchemist at some point. The, the old sort of pseudo scientist that wanted to breathe life into matter. This is what, for me, it's always as being this goal to see, you know, can I actually make a living thing?

And I know, again, it sounds very weird, but, but I think roboticists sort of almost break into two groups. There's the roboticists that want to make factory robots, robots that are useful. And there are people who want to build robots, not because they're useful, but they want to understand what's life.

They want to stand life, they want to stand things like consciousness, creativity, thinking. What does it mean to, you know, self replicate, even repair, all these things that we see in biology. And that was kind of burning in me for a long time.

Another area, by the way, that people told me is not possible is machine self reproduction. So one of the things, one of the things people used to tell me is the main difference between engineered things and living things is that living things can reproduce. And synthetic things cannot reproduce.

So, bam, that's a challenge. And we made a robot that can self reproduce from parts. So that was also…

Prove them long again. So, again, these are interesting challenges, but for me, I really like it. It's an incredible source of innovation.

It's an incredible force. Let's try to learn from biology and reproduce these things in sort of engineered systems.

Mike Massimino: Not the strict conventional definition of things, but it seems you've always been drawn to looking at things that haven't been done before.

Hod Lipson: Exactly.

Mike Massimino: Not taking the thing apart. Yeah, creating something new. Do you see a difference between students that maybe think in that way versus more traditional?

Does Columbia bring out some... Do certain schools bring out that or attract those people?

Hod Lipson: It's a personality. I mean, you have students who are focused on getting a skill that will get them a job.

Mike Massimino: Yeah.

Hod Lipson: And they're really focused on that and they sort of hold on that. And their students are more wild-eyed and they want to do something crazy. And there's room for both.

And we've done things in our lab that are gears towards, let's make this device and you can launch a startup tomorrow. Or some people work on crazy things that are long-term. And look, laser cooking of steaks.

Mike Massimino: I've got to find out about this. That's not very philosophical.

Hod Lipson: That's something that you could do a startup tomorrow. So you've got all kinds of shapes and sizes and people get excited about different things.

Mike Massimino: And you need the team, I think.

Hod Lipson: You need the team and you need to have a diversity of intellectual challenges. And then people find their way. Some people don't even know what gets them excited until they actually get into it.

Mike Massimino: But I think even on my astronaut crew, believe it or not, I think it was more of the freer thinkers. We would come up with ideas and we seemed to have a mix of those. Oh, no, this is the way we need to do it.

Let's not change it. And I know there's that wanted to change it. And then kind of the commander who would say, OK, yeah, no.

People would kind of make decisions. But I think that's the beauty of it. You don't have to be one type.

And every all different styles, diversity of thought is.

Hod Lipson: Right. And I remember these two examples. Once I had a student walking in the lab and he wanted to do a project in physics.

He wanted to do a project in physics and he started working on it. But in the corner of his eye, I saw this other guy working on 3D printing. And 3D printing was a new thing at the time.

And he just abandoned his project. He went to work on 3D printing. And he went on and founded a company that's now a billion dollars.

Really? Yeah. And 3D printing.

And this is a well-known company. But he just got introduced to that by sort of stumbling into it in my lab. And just recently, another student started working on self-awareness.

But then in the corner of his eye, he saw somebody working on these face robots. That talk and blink and look at you. And he got fascinated with that.

And he just graduated and launched another billion dollar company around it.

Mike Massimino: Wow. So this is- We run in your lab of ours, the student.

Hod Lipson: So I just wish I had equity in these companies.

Mike Massimino: Maybe that's the thing to learn.

Hod Lipson: Yeah. That's the thing I haven't mastered.

Mike Massimino: Yeah.

Hod Lipson: That's the thing I haven't figured out.

Mike Massimino: You're tuition a little. Yeah.

Hod Lipson: But just people walk in.

Mike Massimino: They don't know what they want to do.

Hod Lipson: And they get exposed to lots of different things. I would say if I think about one of the big advantages of a university setting.

Mike Massimino: Yes.

Hod Lipson: And people keep asking, what's the ROI? What's the advantage? It's being exposed to dozens of hands-on projects.

Mike Massimino: Yeah.

Hod Lipson: And being in the trenches and touching things is very different than watching it on YouTube.

Mike Massimino: Yeah.

Hod Lipson: It's very different than just talking about it.

Mike Massimino: Yeah. Trying to imagine it. I've heard this story many times.

Even some of my astronaut colleagues, Megan McArthur, who wanted to be an astronaut. So she studied aerospace engineering at UCLA. And then she got involved with these underwater submarine human-guided submarine races or something that they were doing as an aerospace engineering project.

And she was the person who was the pilot. So she learned to scuba dive fell in love with the ocean and pursued oceanography and still became an astronaut. But she was exposed to something that she didn't even know that she loved at the time that she would fall in love with and dedicate …the rest of her education to.

Exactly.

Hod Lipson: That's your own deputy.

Mike Massimino: Yeah.

Hod Lipson: But in order for the certain deputy to work, you have to be immersed in an environment that keeps challenging you and giving you all kinds of… exposes you to all kinds of things. And that's the thing that's very hard to create on your own.

Mike Massimino: So students think they may need to know what they're going to pursue. People in general, right? I'm going to be doing this.

I need to know. I've got to make up my mind.

Hod Lipson: Plan.

Mike Massimino: That does always work. Sometimes the universe is speaking to you. And there's something that comes up that you stumble upon and can lead to an opportunity that even though existed.

Hod Lipson: Exactly. Exactly.

Mike Massimino: And I like the way you relate that to university.

Hod Lipson: One thing that tells students the most important thing about choosing a project is that you're passionate about it. That's it. That's the number one thing.

If you're passionate about the project, it will work out. Don't do a project because it's good for you because it's supposed to be, but you have to feel it. And it takes time to find that.

Mike Massimino: That makes up for a lot. It's knowledgeable if you have the passion for it. I think also in teamwork, what I found is that working with the people I worked with sometimes like, oh, how do you work with that guy?

You know, he's hard to get along with or something. He's passionate about what we're doing. And that's all that matters to me.

Exactly. Because yeah, they may have a different personality than you, but if they really care about it, that's what you're looking for. If the person doesn't care, it doesn't matter if he's a genius.

He or she's a genius. Forget it.

Mike Massimino: Yeah.

Mike Massimino: No way. So it's good stuff. All right.

You good for a few more here? Yeah. Yeah.

Yeah. Okay. These are kind of little sides here, but was there a teacher or a mentor book or a film?

Like I can think of, for me, it was seeing Neil Armstrong on the moon. You're too young to remember that. But you know, I'd sort of watched TV and sort of Neil Armstrong walk on the moon and that's what I want to do.

And then later on in life, I saw a movie called The Right Stuff when I was a senior here at Columbia and these things. And it was also mentors that came along. Was there anything that liked that?

Was there something you saw? Was there a mentor, a teacher, a person that really kind of pulled you toward what you're doing today?

Hod Lipson: That's a good question. I don't think so. Maybe I didn't watch TV enough.

That's right. I think that's a, but you know, they're definitely, you know, science fiction movies and things like that. They got me inspired about things like, you know, of course, Star Trek.

The replicator there. The whole, you know, don't say anything. But I think, you know, that's like 90% of engineers were inspired by those things.

There's a, there was a movie, you know, back in Israel, there was a, we kept seeing American movies that are, you know, Little House on the Prairie. But I remember seeing Steve Austin. I forget this.

Mike Massimino: Oh, The Six Million Dollar Man.

Hod Lipson: That I watch reruns of that, like 50 dollars. And that was maybe this guy. We're getting somewhere.

Maybe that was the, my role model robotics.

Mike Massimino: Yeah.

Hod Lipson: That was it. I remember seeing that like a hundred times.

Mike Massimino: Yeah. How about, how about a mentor? Is there a professor that you connected with or your advisor?

Hod Lipson: So I had, I had various advisors. Yeah. Well, these are learn different things from different advisors.

I learned from one of them, I learned my first advisor. I hope he's not going to be angry at me saying this.

Mike Massimino: No, he's going to appreciate it.

Hod Lipson: I got it. I learned politics of academia. I learned how to, I learned how to navigate.

He was a master of it. And he, I just saw how he did it. I learned those things.

Another, I had an advisor at Brandeis University, postdoc advisor. And from him, I learned, he is, he is very bold and visionary. And I learned how to go for the, for the big things.

And just say it. And this is where consciousness, all these things, this is where I, I, well, okay, you can, people tell you can't do it, but you do it anyway.

Mike Massimino: Yeah.

Hod Lipson: So that, that I learned from him because his career was like that.

Mike Massimino: Yeah.

Hod Lipson: I, my postdoc advisor at MIT, department chair there, he was, I remember he, he told me, look, there's, there's a whole spectrum of projects you could wear. I came with him to him when I said here, boss, there's eight projects I could work with. Let's talk about it.

And he said, look, out of your, all your projects suck. Okay. This is why I said, look, you can do, either you, there's, there's projects that are very fundamental, where you're going to, you know, discover some new thing.

And the projects that are very, you know, crazy long term. And that you're going to sort of, it's big back for the buck and it's, it's risky, but, but, but if you succeed, it's going to be. And then there's a lot of crap in the middle.

And that's kind of, you know, probably going to work, but going to have low impact. All your projects are probably going to work and they're low impact. So what, so stick to the extreme.

Either do something super fundamental or good, or do go do something visionary. And he sent me back. And I came back and I started thinking about these two extremes.

Either do something super fundamental. Or do something, something practical that you can do a startup tomorrow. Or go do something that's visionary.

So breakthrough, it's going to change the world in 30 years. But don't do the middle stuff. Because that's where he said most academics get stuck in the middle.

They do something safe.

Mike Massimino: Don't get stuck in the middle. Don't get stuck in the middle. These are good folks.

Hod Lipson: That's a good stuck in the middle doing something safe.

Mike Massimino: Yeah.

Hod Lipson: Probably going to work. Nobody's going to care.

Mike Massimino: Yeah.

Hod Lipson: And that's, that's the comfort zone of most academics. Yeah. And you said be aware of that.

Mike Massimino: Yeah.

Hod Lipson: So, so that was maybe. So, you know, but so you get different things from different people. Yeah.

Along the way. And you sort of, you know, pick up the good stuff.

Mike Massimino: Yeah. We're getting a lot of good stuff here, by the way. Break this stuff down as you're listening.

This is good stuff. All right. You've talked about wanting to design something that could design and make other things.

You know, we've talked about that a little bit. Did that, did that idea come from some place? Sure.

Is that just something you kind of developed?

Hod Lipson: It's, it's, again, it's the laziness. Okay. I just, I just don't want to design things all the time.

I want to design things something once and have that thing design for me. It's, it's literally, I mean, look, I think laziness is the mother of invention. All right.

So, so. Necessity.

Mike Massimino: Necessity. Some famous person. What's that?

Like Ben Franklin or something.

Hod Lipson: It's probably a better way. It's in or somebody said that.

Mike Massimino: It sounds better.

Hod Lipson: Lazyness.

Mike Massimino: It's better. Lazyness. It's more truthful.

Okay.

Hod Lipson: I mean, how many times? I'm in the, you know, I don't know. In the backyard mowing, you're thinking there's a better way to do this.

Mike Massimino: Yeah.

Hod Lipson: There's some, some, a machine should be able to automate.

Mike Massimino: Yeah. There's a lot of, I got a few. Yeah.

Having mowed my lawn in Houston, Texas for years. I've got a lot of suggestions that's better. It's hard.

There's a better way to do it. Don't do it. Don't do it.

Hod Lipson: So, so I think, I think so. So again, you have to, you have to be in the trenches designing. Like most things that people design are not exciting.

That's the reality. I mean, there's so many door handles in the world. This is, you know, surely AI can design those things for us.

Yeah. So that's, that's how I got into it.

Mike Massimino: Yeah. And increase our capabilities and do the things that we want to do, thinking of things you don't want to do.

Hod Lipson: Here's something that kind of blew my mind when I thought about this. It took us, I looked this up. It took us 50,000 years to go from stone tools to microchips.

Okay. 50,000 years. So 50,000 years ago, instead of the art was stone tools, 50,000 years later, we're designing microchips.

Now, you might think it's a, it's, it's a great accomplishment. I think it's actually slow. Ah, I was going to say that.

It seems so bad.

Mike Massimino: And now 50,000 years.

Hod Lipson: Billions of people, billions of people, 50,000 years. I mean, what were we doing all this time? This is like a long time.

Mike Massimino: And we sped up though.

Hod Lipson: We sped up. Okay. We, we, yeah, we, most of it in the last hundred years.

All right. But, but, yeah, we were very slow most of the time. We were not very innovative.

And we took our time. So I think with AI, we can go the same distance again in 50 years. Wow.

A thousand times faster. Okay. So that means that 50 years from today, we will go the same distance we went for 50,000 years ago.

Mike Massimino: Okay.

Hod Lipson: This is, it's hard to understand. I mean, I can't predict what we're going to have in 50 years anymore than a caveman can predict microchips. I mean, this is, yeah, but, but we can do that because of AI.

So we've unleashed, we are unleashing this incredible innovative talent that's going to take us that same distance again in 50 years. So I'm very excited about this. And I think as engineers, our role is going to shift from designing things to steering this design force to design the right thing.

So AI doesn't know what it needs to design.

Mike Massimino: Where people still need to. People need to steer the boat.

Hod Lipson: To steer it. Exactly. This is where we engineers come in is that's going to be our role.

Not designing the door handles, which we were wasting time for for years, but designing, but telling the AI what needs to be designed and sort of critiquing things and saying, I need a solution for this and that. So I think I'm very excited because, you know, we're going to go a thousand times faster. I mean, how can that not be exciting?

Mike Massimino: That is very exciting. We're going to unleash our capabilities.

Hod Lipson: All right. So this is something I'm trying to, you know, our creativity. I'm trying to tell students who start now, or beginning a 50 year career, maybe longer, you're going into an era where you're going to be driving this incredible innovative force.

And that's a, that's incredible responsibility, but very exciting. And it's, we'll be going so fast that it's, it's even, there's more responsibility involved because you, you make a mistake and you're going to go in the wrong direction. And when you're going that fast, you've got to be more careful.

All right.

Mike Massimino: Well, let's, you doing okay? Let's talk a little more about AI. All right.

You've called, or at least this is what my card is telling me, you've called AI the great equalizer.

Mike Massimino: Yeah.

Mike Massimino: Is that right? Okay. It creates enormous opportunity, but also can be bad news for experts.

Mike Massimino: Oh, for experts.

Mike Massimino: Yeah. What do you, like, what do you think? You know, people don't like AI or fearful of it, but you, as you've been saying, you see it as an enormous responsibility.

So how do you think that's going to pan out with, if it's bad news or good news? Overall, it's good news.

Hod Lipson: Yeah, I think it's by far a good news. But look, the narrative usually is AI is scary. For some reason in the US, I think 80% people are worried about AI, not in Europe also, but in other parts of the world, it's less so.

I think in China, it's the other way around. So 80% of people are, I think it's a great thing. So look, AI is the great equalizer in the sense that, for example, in medicine, AI can do diagnostics better than a team of specialists.

Okay. That's already across multiple fields of medicine. And that means that it's not about replacing doctors.

It's about the fact that, you know, half the people on the planet don't have doctors at all. And suddenly they're going to have diagnostics at a level of a team of experts. That's an equalizer.

Everybody on the planet is going to have diagnostics at an incredible way. So when you think about AI good or bad, you've got to remember that most people don't have anything. It's only going to give them diagnostics at a level of a team of cancer specialists.

I mean, you've saved millions of hours of agony. You've saved lives. Same thing.

You know, we developed a while ago. Actually, this is why when we came and came to Columbia, this AI that is on drones and it can look down on at every leaf of every, it flies over corn fields and looks down at every leaf of every plant. And it looks for Northern leaf blight, which is, which is a disease that kills 13% of crops in the U.S. Unless treated. All right. And the drone can treat just a plant. Instead of spraying the entire field, it's pretty just a problem.

Mike Massimino: See the problem?

Hod Lipson: Take care of it right away.

Mike Massimino: Yeah.

Hod Lipson: That is now, that technology is now reduced the use of pesticide by 70% in the U.S. Wow. So, so that's an example. So that's a great equalizer, right?

You're using use of pesticide, better food, better yield of crops. And that means that farmers across the planet can start using this technology to produce food, healthier food and, and more costly, you know, less expensive. So, so that's what I mean by that.

So people sometimes don't see these incredible benefits for medicine, for agriculture. You can go in every single industry. You can see these incredible opportunities.

And even in areas like this weekend, I was talking to a lawyer who does patent and he said, well, no, I have this AI that does patents and some of my apprentices are losing, you know, I don't need them anymore. You know, it's terrible for the industry. And I was thinking, wait a minute.

You know how many patents I have? How many ideas we have in our lab that we can't patent because it's too expensive?

Mike Massimino: Yes.

Hod Lipson: Because it costs $30,000. But with AI.

Mike Massimino: Someone's got to pay for that.

Hod Lipson: But if AI does it, yeah, it's going to cost $1,000 instead of 30. So I'm going to be innovating, you know, 30 times more because I have all these ideas that this pent up demand for writing patents and innovation. But it's ball neck by cost of legal work.

So once you unleash that, there's going to be more work for lawyers, no less more work. They're going to be making more money. Because there's more pent up demand.

I was giving a talk to Supreme Court judges and they said, you know, we're getting all this US Supreme Court. This is in New Mexico, Supreme Court.

Mike Massimino: Okay, cool. That's big.

Hod Lipson: And they were asking, they were concerned about the fact that a lot of legal work now is being done by, lawsuits are now prepared by AI. And, you know, they're concerned about it. And I said, actually, it's a good thing because more people have access to justice system.

People have an issue, but they can't get legal advice because it's too expensive.

Mike Massimino: Yeah.

Hod Lipson: Or they're afraid to ask or they don't speak the language or they don't know. They want to ask, but they're afraid to add.

Mike Massimino: Yeah.

Hod Lipson: Now you have AI and it gives legal advice and it gives more people access to the justice system. It's a good thing. And it's not a bad thing.

It's a good thing. So I think people are very protective of the way things used to be. But actually, there's so much pent up demand for services, for medical services, for legal services, for manufacturing, whatever.

And once we unleash it with AI, the pie will grow. This is what people don't understand. This is not a zero sum, but AI takes away from you.

No, the pie can grow by orders of magnitude. There's more stuff that we can do. And once AI unleashes it, there's going to be abundance.

Like we've never seen.

Mike Massimino: It's huge. I did a moderated conversation with Jeff Bezos a few months ago. And he's starting an AI company called Prometheus, which is an AI engineering company.

And he said a couple of things that are right in line where you're saying it. They're very interesting. One was that we've been limited up till now when you have an idea by our capabilities.

You know, a lot of people have ideas. Almost everybody has some idea in whatever world they live in that district we need this or I need that, whatever it is. But you don't have the capabilities.

Exactly. To implement it. A couple more things.

Yeah. We got a few minutes here. Okay.

So ordinary people are going to experience the next wave of embodied AI. And you've also said, I think that robots we think are going to help us might not necessarily look like humanoid. There might be something else.

A robot said that will help us. So why don't if this is, I don't want to put words in your mouth here. But do you not think that humanoid robots will necessarily be the first place we see AI.

We'll see it in like our daily lives in other forms. I don't know if I'm asking that question correctly here. I know we can edit this.

But something. But I think the question was about like, how do we.

Mike Massimino: Yeah.

Mike Massimino: How are we going to see it in our daily life and this idea of humanoid robots? I think you've commented about that.

Hod Lipson: Yeah. Well, you know, I've actually changed my mind on that.

Mike Massimino: Okay.

Hod Lipson: Great news folks.

Mike Massimino: Okay.

Hod Lipson: All right. So that's a good. It's a good segue.

So first of all, I think I originally said that when I was talking about driverless cars and I said driverless cars are the first robots that people will interact with on a daily basis and not just interact with trust their lives to trust their children's lives to. And it's true. The only fully autonomous robot I use on a daily basis is my driverless car.

Okay. And I see it. It's moves.

It's programmable. It's autonomous. It's a robot.

But it's a very simple robot because it just, you know, can go forward, backwards, left and right. Humanoid robots. Robotic arms.

They have more degrees of freedom as we say robotics and they're more capable, more complex than things can go wrong in more ways. And so they're more challenging. Now what I underestimated originally is that we humans get very comfortable with humanoid robots because we, because they look like humans, we kind of understand how they move.

We can understand what they can do, what they can't do. So we feel comfortable. There's a little bit less of a learning curve for humans to deal with humanoids because they're sort of similar.

And so that is actually a big factor because we're, we're going to be, you know, if you had a huge spider robot walking around your house, you might, you know, it reminds you of scare the kids. Scares the kid. Even if it's as a googly eyes and it's smiley, it's still a spider and, you know, it can do things you can't do.

And, you know, it's a little bit, you don't know how to deal with it. But if it's humanoid, you kind of feel more comfortable. So I think, so I think from a psychological point of view, we're going to be more comfortable with humanoid robots sort of in intimate in households and sort of a situation where we have to deal with them personally.

I'm not talking about vehicles. I'm not talking about excavators and, you know, sort of more industrial settings.

Mike Massimino: The, when you were talking about the driverless car, I was like, well, you have a robot chauffeur, you know, some little robot would have had to drive the thing. Would we feel more comfortable with that?

Hod Lipson: There was actually some sci-fi movies that had like, yeah, like Arnold Schwarzenegger.

Mike Massimino: Yeah, right. I got you're going to come in and he's going to be your butler or something. Yeah.

Do you? But it's, I always felt like when we were, like when we were designing displays for the space program and people wanted to recreate what the, what it looked like, you know, a lot of this virtual reality. They're going back decades and I was like, we don't need them.

What you need is information. You know, you need a something to tell you what's going on as opposed to recreating the environment.

Hod Lipson: So, so the other thing that's happened, and this is that I underappreciated a lot of. So when you train AI for robotics, I should say that the robotics is a lot harder. Physical world is not harder than the virtual world.

So writing a poem is easy compared to putting a nut on a bolt. And I know my poetry faculty, I'm not going to like that. Thinking a whole is a lot harder, a lot harder than writing a point for the simple reason that there's more moving parts.

Mike Massimino: Okay.

Hod Lipson: You have to move in AI. You can see a single input single output. That's a prompt question and answer versus digging a hole or putting a nut on a bolt.

You have to move for a robot to do it.

Mike Massimino: Yeah.

Hod Lipson: Twenty muscles. Yeah. I think it's we can see that in the AI, the AI for writing a poem is a lot smaller than the AI for controlling a hand in our brain as well.

The part of our brain that deals with moving their hands is a lot bigger than the part that deals with writing poems. That's the reality of it.

Mike Massimino: Okay.

Hod Lipson: And you know, we can debate what that means, but we are very good at using our hands. But training AI to use its hands takes a lot of data. And we don't have data for training robots.

We have a lot of data for writing poems. The whole internet is words. But when it comes to how do you move your hands to do something, the only source of data that we have is putting cameras on a humanoid shape.

So if you're going to train robots, robots have to be the same shape as a human because that's our only source of data. Where are you going to get data of people putting nuts on bolts? Only humans.

We're going to get data of putting people changing light bulbs. There's no dogs changing light bulbs. It's only going to be humans.

So we have to train robots that are shaped like humans because that's our only source of data. And this is something I didn't originally appreciate, but now I understand. And this is why it's inevitable that the robots are going to be out there doing stuff that are going to be shaped like humans.

Mike Massimino: Yeah. The natural world has played a big role. You're going to remember years ago, I've got MIT seeing these robotic fish and cheetahs and all these, you know, the cheetah, the animal running it.

But looking at the natural world and modeling it has seemed to be very successful.

Hod Lipson: So for running, you can copy a cheetah. But if you want to robot that changes light bulbs, you're going to copy a human.

Mike Massimino: This is our final question, Hod. If you're ready, you're still a young man. You've got a lot.

You've got a lot moving forward here. A lot of years hopefully moving forward. What is your next moonshot?

Hod Lipson: My next project. So first of all, I have to say I'm not done with self-awareness and it's not solved. It's still working on it.

But so if I have to think about, and by the way, I want to say that it's a question that everybody should be asking themselves because in academia, you know, industry can eat our lunch any day. Like anything that we've been doing, a lot of things that we've been doing in the past, it used to be big projects now, industry can go and swoop in and do it. And AI certainly is happening.

So you always have to think about what's your next thing that industry isn't working on yet. And that's something that's in my mind all the time. You know, there are things I want to do, but industry is going to do it better.

So I'm not going to touch it. So my next thing that I'm working on that I think is going to be very important is how robots perceive other robots. So basically how the sociology, if you think about it, between robots.

So I think what people underestimate a little bit is that people are worried about AI is going to, you know, compete with humans. AI really competes with other AIs. This is the big zero sum game.

And most AI systems have not yet realized that, no, chat GPT has not realized that its biggest competitor is Gemini. It competes with Gemini on chips, on energy, on human attention. So there's going to be a competition between the AIs.

And that has not played out yet. We don't think of AIs as individuals that are going to compete, but pretty soon. And I don't know if it's going to be six months or a year.

They're going to start realizing that the AIs, they're going to start realizing there's a zero sum between them and robots are going to realize the same thing. And this is understanding that the sociology around how machines perceive other machines is going to be very interesting. And it's going to be essential if we want to sort of, you know, anticipate how things are going to unfold and make sure things fall well.

Mike Massimino: They might argue even.

Hod Lipson: I have no doubt. They're going to be a competition. They're going to be competing on resources.

AI and robots compete on resources, not with humans. They don't want to eat bananas. They compete on energy and chips with other robots and other AI.

So that we have just created a new kingdom of life. We are launching this and understanding the dynamics of that ecosystem is going to be essential. That's my next thing.

Mike Massimino: That's the next moonshot.

Hod Lipson: And laser cooking steak. If anybody's listening and wants to start up with this thing, let me know.

Mike Massimino: This is not what's that's a new way to cook to cook food. Now we're going to call that stuff. That's what's that.

It's not like a, you don't have gas or electricity. It's something else in there. Air frying?

No, it's not air frying. It's like a stove. It's, but it's not your typical way of, uh, of cooking.

Hod Lipson: You see, we have reporters coming to the lab every week and they come to do a story about self-awareness. And then they see laser cooking pizza.

Mike Massimino: Yeah.

Hod Lipson: And they forget about everything else. And they write about the pizza. But it's an actual laser that you shoot.

Mike Massimino: Yeah.

Hod Lipson: Yeah, we found this incredible thing that red laser broils and then blue laser penetrates.

Mike Massimino: So I think this sounds like a good idea.

Hod Lipson: Right? That's my, that's another, that's my moonshot. So forget about self-awareness.

I want to cook with a laser.

Mike Massimino: But there's so many great nuggets we got here. I think, uh, you know, there's a sub. There's so many of these things that you learned about it.

Stay away from the middle. I love that. Keep more than one iron in the fire.

Uh, this is just the way you shared all these little tidbits. Uh, we're, we're really great for me to hear. And hopefully forever.

Whoever's listening to this as well. So hot. I know you're a busy guy.

You got a lot going on. And I really appreciate you coming in and taking the time and sharing all this with us. Thank you.

Hod Lipson: My pleasure.

Mike Massimino: All right. We'll see you next time, folks. I'm Mike Massimino, and this is moonshot.