Azeem Azhar

Or is this something where Congress in particular needs to come in with a particular point of view?

I mean, a few days ago, President Trump said that

AI training is akin to reading a book or an article and many people who are doing research and writing say, well, that's not the case because it's used in a very, very different way.

I'm just curious about how you think we equilibrate, right?

We get to this new equilibrium and what help we need from Congress or the president or anyone else.

Let's go back to that point about machine scale.

One of the things that my team and I calculated was that we reckoned that at some point this summer, more tokens, which represent about three quarters of a word, were produced by machines talking to humans or to other machines than are produced by the entirety of humanity.

And of course, that wasn't the case five years ago.

And by the winter, they will be far ahead of us.

And you talked about how internet traffic, which has a lot of bots right now, will be vastly dominated by bots.

And part of that will be this idea of the agent, thousands of agents that each of us will have across this agentic web going back and forward.

And it looks like it's a tremendous market expansion because those agents will grow.

all require information.

Some of it will be private transactional information.

Other of it will be telemetry signals, environmental information.

So how does this affect the way that we can think about the future of that agentic web?

I mean, prima facie, it seems like it's going to be a much larger market, but transactions will have to be much smaller.

Quite often, really, really important and valuable information is not going to get bought at large rates by large numbers of people.

And you could look at academic research in that vein.

And what's happened with academic publishing over the last 25 years is that a couple of academic publishers have started to absolutely dominate that market.