The Race to Recursive Self-Improving AI

Dhaval Adjodah (0:00)

The fundamental idea here is to build AI that improves itself, and I think we're getting there.

Owen He (0:05)

For the first time, we could swap the optimizer as the optimizer, and this leads to a reinvestment of cognitive improvement into itself.

Dhaval Adjodah (0:15)

You can think of RSI at many different levels of the stack. Prompts, harness, weights, you can go all the way to basically chips. Our goal here is to optimize all the levels of the stack.

Owen He (0:27)

Our approach to RSI is that we have a multi-agent system working together that's been recursively improved.

Martin Casado (0:34)

I'm so impressed by your belief that this idea of a model improving itself.

Dhaval Adjodah (0:38)

But it's not a belief.

Martin Casado (0:39)

That is a dramatic statement, which is I think it's the first time I've heard you say it.

Dhaval Adjodah (0:42)

Because we have been so thorough, we have not released things until we're super sure it's correct.

Martin Casado (0:47)

Do you envision a system where the amount of energy that goes into it somehow decreases over time relative to the improvement?

Dhaval Adjodah (0:55)

I can't share too much, but one of the very important things that we've tried to do.

SPEAKER_4 (1:00)

A16z general partner Martin Casado sits down with MakerMaker.AI co-founders Dhaval Adjodah and Owen He to unpack recursive self-improvement, or RSI, and separate the increasingly popular term from what it would actually mean for an AI system to improve itself.

Dhaval and Owen distinguished true RSI from simply using AI to automate pieces of AI development. Their goal is a closed loop, where the system being improved is also doing the improving across components including models, harnesses, and eventually more of the AI stack.

They explain why closing that loop still doesn't guarantee runaway takeoff, and why measuring real progress requires accounting for both capability gains and the resources required to achieve them. They also share what they're seen in their own research, including early signs that self-improving systems can increase performance while improving efficiency. And they dig into the practical challenges of proving those gains are real, from variance and reward hacking, to the enormous difficulty of distinguishing genuine improvement from noise.

Martin Casado (2:04)

Welcome both of you. We're going to be talking about everybody's favorite topic, which is RSI, recursive self-improvement. Maybe to start, it'd be great to just introduce yourselves and your backgrounds, and we'll talk a little bit about what you're working on now, and then we'll dig into the content.

Dhaval Adjodah (2:17)

I think of myself as a physicist of some kind. I studied that, then I did a PhD in reinforcement learning at the MIT Media Lab.

I really like building things, and I'm very excited to be building a cool startup, Maker Maker, with Owen and Grace.

Owen He (2:32)

My name is Owen. I was first studying math, and then one day I realized that all mathematicians, too, are proving theorems, and I was taking a class called Formal Language and Logic, and they were showing how you can build a theorem prover. I was like, oh, that would be cool. I can just build a machine that prove theorems for me.

That's how I got into AI.

Dhaval Adjodah (2:51)

Amazing.

Owen He (2:52)

And then for my PhD, I work on continual learning and meta-learning. Then worked for a few years at DeepMind as a research scientist.

Now I'm working on RSI.

Martin Casado (3:02)

I see a lot of pictures come in, and a lot of them use the term RSI.

And I'll be like, oh, so what do you do? And they'll say, we use AI to create AI. That sounds pretty interesting. So what do you do? We use AI to create GPU kernels, which that doesn't, to me, kind of met the definition of what I would consider to be RSI. So maybe can we start by you defining what you think or what you would include in RSI definitionally?

Owen He (3:25)

Yeah, I think RSI is definitely an overloaded term today. Many people when they say RSI, they mean different things. So I think the important thing when you are differentiating them is to ask the question, what is the optimizer and what is the optimizer? When you're self-improving, who is improving? So there are people who are saying, OK, we're doing RSI. We have humans using AI to develop better AI.

But that to us is just auto-catalysis, in a sense that the improver is still human with tools, even though these tools are getting better and better, but the ideas and the improvement come from human brains.

Martin Casado (4:01)

Right. I don't mean to interrupt, but to be clear, most technical epochs have had auto-catalytic effects. There's no way you can create a computer chip without a computer. You could argue that the Internet development was sped up by the fact that people that weren't geographically located could coordinate, right? I mean, the steam engine, you could build better steam engines because you get coal faster, etc. So this notion of using the thing to build a thing faster is not new.