That raises a huge philosophical question: What actually counts as consciousness? And should AI get rights?

Some are saying we're effectively at the stage where in the right harness, models can be considered conscious.

But I think personhood should not be granted by capability.

It should be by origin.

As Imad argued at Oxford Union, the right relationship between humans and artificial minds is a treaty, not enrollment.

Now, switching gears to OpenAI's Astra, this feels like a genuine turning point for mathematics and everything downstream of it.

Math too cheap to meter, right? The phrase captures it perfectly.

Astra is solving problems that have been stuck for decades.

And how do people get good at that? What, what did those look like?

Yeah, I mean, I think, uh, it's...

You- you'll have a lot more success when working with your virtual agents if you can clearly specify what it is you want.

And the clearer you are on what it is you want, the more the agent will have sort of guidelines and sort of rules of, you know, an outline of what it is trying to accomplish.

Um, whereas if you don't specify very much stuff, the agent has to sort of infer what it is you meant, and in many cases it might infer things that are different than what you imagined.

So we've always told computer scientists from the very beginning that really, it's really important to specify what it i- what's the software that you're writing is trying to accomplish before then going and writing it.

And so now we actually have agent-based systems that can do the writing, but the importance of specifying what you, what it is you want has actually gone up because before you'd be handing it off to a very intelligent human who maybe has context or can ask you follow-up questions.

What, what would you do with the cur- the Jeff Dean today, 25-year-old Jeff Dean?

[laughs]

I mean, I can weigh in.

I, I think, you know, that was actually pretty obvious, that if you have a model that is used by a lot of people, that you're gonna get a lot of lessons and experience of what's working well, what's not working well.

And, you know, we've seen this in Search for many, many years, is the, the use of Search by our users really informs, you know, what are the things it's not doing well, what are the things we should be doing better.

Um, you know, aggregating lots of interesting usage statistics to understand, you know, that i- in a, in a deeper way, and then working on improving those things is important.

And AI models should be no different, right?

Better to have one model than five little ones.

One thing I would say is like from the start, we wanted Gemini to be-- I mean, even pre-Gemini, one, one of the origins of the Pathways project was to explore, you know, a single model that could do many things, a multimodal model that could deal with all different modalities, a very large model that was sparse, so you would activate different pieces of it for different kinds of things.

I had the awesome luck to chat with Jeff Dean, chief scientist of, of Google for the past however years, and I wanna show you the video of me asking him why Omni is different

Well, in the same way that Nano Banana suddenly brought a Gemini model to image generation and let you interactively edit that image with-

Yeah

... uh, and have a picture of your dog or whatever.

The same thing i- is what Omni brings to video.

So now you can generate video with a text prompt or images and say, "Please make these people walking down the street," or whatever, but you can also interactively edit the video-

Mm

... information.

And I think an LLM-based system is not gonna be that dissimilar, right? You're gonna attend to trillions of tokens, but you're gonna want to identify, you know, what are the thirty thousand-ish documents that are with the, you know, uh, maybe thirty million interesting tokens.

And then how do you go from that into what are the hundred and seventeen documents I really should be paying attention to in order to carry out the task that the user has asked me to do? Um, and I think, you know, you can ima-- you can imagine systems where you have, you know, a lot of, uh, highly parallel processing to identify those initial thirty thousand candidates, maybe with very lightweight kinds of models.

Um, then you have some system that sort of helps you narrow down from thirty thousand to the hundred and seventeen, uh, with maybe a little bit more sophisticated, um, model, uh, or a set of models, and then maybe the final model is the thing that looks at the hundred and seventeen things that might be your most capable model.

So I think it has to-- it's going to be some system like that, that is really enables you to give the illusion of attending to trillions of tokens.

Um, sort of the way Google Search gives you, you know, not the illusion, but you are searching the internet-

Mm