OpenAI 向 10 万名学术研究者免费开放前沿模型

What I tell my kids is what I do at work is I contemplate the mysteries of life, the universe, and everything. Try to figure out, you know, where we came from and where we're going.

[music]

I'm trying to figure [music] out how to do physics in the new year. I feel like AI is erasing the boundaries between subfields of physics. So, basically I'm trying to figure out what I can do with the model, not just [music] trying to do better of what I've been doing before.

I'm a physicist, a cosmologist, and also an AI researcher. I'm here working on all of all of the above. We were trying to predict signatures of the microwave background light that we see in the universe. And the model made a great analogy to research in photonics. And [music] that actually helped us get unstuck on a particular piece of the problem.

I've been using um

[music]

AI a lot, mostly as a way to quickly generate code that can analyze data and test various hypotheses very quickly. So, you can get a rough feeling for where the signal might be in the data.

And so, I just gave it my paper and like [music] 10 minutes later it had histograms basically for every tissue. And it was really striking. The metric that we had developed was really high and strong in the brain and lower in other tissues, which was really compelling. It's just [music] really fun to basically make a discovery by using plain language and just talking to the computer. [music]

The breakthroughs are, you know, very often not by solving a very hard problem by making the right assumption. [music] I'm starting to have the AI model bring me like new conjectures that I haven't thought about before and can write numerical tests and, you know, there there are many ways you can you can you can test whether the ideas are are correct, but [music] just to have the idea to begin with uh is uh you know, it's such a big deal.

In many of these problems where what it really is going to take is not vast scale of the experiment, it's going to take sort of extreme [music] cleverness or a great leap of insight. Our ability to make these sorts of, you know, leaps of experimental imagination is just limited by the number of people who are able to make that leap. With AI it can be just a huge force multiplier. If any one of those experimentalists now has all of their AI agents, then the space of ideas, [music] the number of leaps of insights they can have, that goes way up.

Suddenly we'll be able to make progress on things in the scale of a year or two years that we thought was at the scale of decades.

[music]

Things that were a lifetime away, maybe now will become a decade away.

I honestly have had so much fun with AI at examples like that. It has freed me up to spend more time on the stuff I love about my job, which is the scientific discovery, reading other people's papers, meeting with my students. [music] So, I think just removing the energy barrier to getting started with it is going to be really huge for the scientific community.

These tools absolutely need to be widely distributed. [music] It's essential for the progress of of science.

Scientists anywhere in the world, any individual [music] could be the one who leads the next breakthrough in whatever field they work in.

The impact could be absolutely enormous, and I'm really excited to see what comes of it. People will use it in all sorts of creative ways, [music] and we it's not that we can sit here and predict what that will look like, but it's surely going to be an accelerant. [music]

That curiosity, the thing that got us into science I think in the first place, to have that much more present in our day-to-day [music] professional lives has just been awesome.

I mean, scientists are very creative. It'll be really interesting to see what scientists come up with once these tools are in more hands and more broadly used. [music] I think it's a great community to look for new uses of AI.