Ahmed El-Kishky

Maybe also there's a new class of, like, problems that we don't even know about.

Maybe, you know, people can, you know, modify their program with AI to sort of, like, compete against each other.

So that way you incorporate AI somewhere.

I think it's hard to know where the world's heading, but I...

really believe there will continue to be these competitions where students and individuals are pushing the limits of what humans can do and sort of improving.

The thing that actually interests me the most is pushing human knowledge.

I think it'd be really great if our models were contributing to humanity.

I think we've...

kind of pushed to the limits of competition.

I think the competitions are not really an end goal in themselves.

There is sort of a way to benchmark progress.

If we're able to sort of improve on these, that's great.

But we don't want to train AI just to play and compete in competitive programming, competitive math.

We want to make sure our models are smart enough to do things like be a valuable, for example, coding assistant to software engineers, be valuable engineering partners to other forms of engineering.

We want them to be an aid to scientists, academics.

Something that's really interesting to me is, yeah, can our models actually discover some new knowledge?

It can maybe be a...

Maybe a new algorithm that no human has ever discovered or a math proof that's stumped mathematicians for a long time or human biology, something in chemistry.

All of these, I think, are things that I think would be great if the models could contribute to.

And that's a little bit where I see things going.