Dario Amodei

Like I could point you to a hundred people here who are better, who are better at that than I am. Um, but, but the, the thing that, that, that I think I did have that was different was that I was just willing to look at something with new eyes, right? People said, oh, you know, we don't have the right algorithms yet. We haven't come up with the right, the right way to do things.

And I was just like, oh, I don't know. Like, This neural net has 30 million parameters. What if we gave it 50 million instead? Let's plot some graphs. That basic scientific mindset of, oh man, I see some variable that I could change. What happens when it changes? Let's try these different things and create a graph. This was the simplest thing in the world.

Change the number of... This wasn't like... PhD level experimental design. This was like simple and stupid. Like anyone could have done this if you just told them that it was important. It's also not hard to understand. You didn't need to be brilliant to come up with this.

But you put the two things together and some tiny number of people, some single digit number of people have driven forward the whole field by realizing this.

uh and and it's you know it's often like that if you look back at the discovery you know the discoveries in in history they're they're often like that and so this this open-mindedness and this willingness to see with new eyes that often comes from being newer to the field often experience is a disadvantage for this that is the most important thing it's very hard to look for and test for but i think i think it's the most important thing because when you when you find something

some really new way of thinking about things. When you have the initiative to do that, it's absolutely transformative.

It's another example of this. Like some of the early work in mechanistic interpretability, so simple. It's just no one thought to care about this question before.

I think my number one piece of advice is to just start playing with the models. This was actually, I worry a little. This seems like obvious advice now. I think three years ago, it wasn't obvious. And people started by, oh, let me read the latest reinforcement learning paper. Let me kind of... No, I mean, that was really the... I mean, you should do that as well.

But now, with wider availability of models and APIs, people are doing this more. But I think... I think just experiential knowledge. These models are new artifacts that no one really understands. And so getting experience playing with them. I would also say, again, in line with the, like, do something new, think in some new direction. Like, there are all these things that haven't been explored.

Like, for example, mechanistic interpretability is still very new. It's probably better to work on that than it is to work on new model architectures because it's, you know, it's more popular than it was before. There are probably like a hundred people working on it, but there aren't like 10,000 people working on it. And it's, it's just this, this, this fertile area for study. Like, like it.

You know, there's so much like low-hanging fruit. You can just walk by and, you know, you can just walk by and you can pick things. And the only reason, for whatever reason, people aren't interested in it enough. I think there are some things around... long, long horizon learning and long horizon tasks where there's a lot to be done.

I think evaluations are still, we're still very early in our ability to study evaluations, particularly for dynamic systems acting in the world. I think there's some stuff around multi-agent, um, skate where the puck is going is my, is my advice. And you don't have to be brilliant to think of it.

Like all the things that are going to be exciting in five years, like in people even mentioned them as like, you know, conventional wisdom, but like, it's, it's just somehow there's this barrier that people don't, people don't double down as much as they could, or they're afraid to do something. That's not the popular thing.

I don't know why it happens, but like getting over that barrier is that's the, my number one piece of advice. Yeah.

Yeah.

Yeah. Um, I mean, uh, so first of all, we're not perfectly able to measure that ourselves. Um, uh, you know, when you see some, some great character ability, sometimes it's hard to tell whether it came from pre-training or post-training, uh, we've developed ways to try and distinguish between those two, but they're not perfect.

You know, the second thing I would say is, you know, it's when there is an advantage and I think we've been pretty good at in general, in general at RL, perhaps, perhaps the best, although, although I don't know, cause I don't see what goes on inside other companies. Uh, Usually it isn't, oh my God, we have this secret magic method that others don't have, right?

Usually it's like, well, you know, we got better at the infrastructure so we could run it for longer, or, you know, we were able to get higher quality data, or we were able to filter our data better, or we were able to, you know, combine these methods in practice. It's usually some boring matter of kind of practice and tradecraft, right?

Um, so, you know, when I think about how to do something special in terms of how we train these models, both pre-training, but even more so post-training, um, you know, I, I really think of it a little more again as like designing airplanes or cars. Like, you know, it's not just like, oh man, I have the blueprint.

Like maybe that makes you make the next airplane, but like there's some, there's some cultural trade craft of how we think about the design process that I think is more important than, than, you know, than, than any particular gizmo we're able to invent.