Technology

If you tried to achieve that purely through informal methods—where hallucination is a persistent risk—getting to the same level of consistent correctness would likely require a lot more research effort and resources.

A model that's really strong at mathematical reasoning is likely to be strong at coding. And a model that's excellent at both math and code is often very good at analysing the nuts and bolts of legal reasoning as well.

What we envision is humans using informal reasoning and intuition as a powerful guide, with formal systems then verifying those ideas. In this way, the formal system grounds high-level intuition.

We see math as code and code as math. The real magic, and the key transition, comes from combining AI, programming languages, and mathematics—bringing all three pillars together.

You feel like a product when you're on these things. You feel like you're selling yourself. But what's amazing is that when you get people in person, actually interacting, much of the time they don't do that same optimisation, especially if you give them just a few nudges.

You cannot expend more energy than you can consume. That's a fundamental law of physics. If you do, you starve, you die, and you remove yourself from the gene pool. Biological systems have therefore been under enormous selective pressure to develop highly efficient intelligence.

The things that are trivially easy for humans turn out to be extraordinarily difficult for machines, and vice versa. I cannot do the square root of a large number in my head, but my pocket calculator can do that instantly. But my pocket calculator still cannot make me a cup of coffee.

The brain is definitely not doing computation in the purest sense. We are not crunching numbers in binary ones and zeros in our heads. A more important question is: what are your inputs, what output do you want, and how intelligently can the system get from one to the other?