As AI models become more capable and intelligent, they can reason through nuanced issues well on their own, so the right approach is to allow closer access to the raw model rather than having humans hard-code rules on top.

Sundar explains that Gemini 2.5's improvement in handling sensitive topics came naturally from the model's increased intelligence and reasoning capability, rather than from hard-coded rules — and that the scientific approach is to let models reason about nuance from the ground up. ✦ AI generated

Sundar Pichai · Lex Fridman · 2025-06-05 · original ↗

I think one of the good insights here has been, as the models are getting more capable, the models are really good at this stuff, right? And so I think in some ways, maybe a year ago, the models weren't fully there. So they would also do stupid things more often. And so, you know, you're trying to handle those edge cases, but then you make a mistake in how you handle those edge cases and it compounds. But I think with 2.5, what we particularly found is once the models across a certain level of intelligence and sophistication, they are able to reason through these nuanced issues pretty well. And I think users really want that, right? Like, you want as much access to the raw model as possible, right? ... from a scientific standpoint, like making sure the models, and I'm saying scientific in the sense of like how you would approach math or physics or something like that, from first principles, having the models reason about the world, be nuanced, et cetera, from the ground up, is the right way to build these things, right? Not like some subset of humans kind of hard coding things on top of it.

verbatim transcript · starts at 36:40