Arthur Kroeber

But the specific problem that we were dealing with there were these very self-contained systems that were clearly under highly centralized control by a couple of people.

And so having that one-to-one line of communication was an effective response to that particular problem.

With AI...

It's much more diffuse, much more decentralized, much more unpredictable what the problems could be.

And, you know, it's kind of easy, I think, to create – easy, we say in retrospect.

It was hard at the time, but it was a –

sort of, if you will, maybe a two or three-dimensional process to come up with game theory to enable us to manage the nuclear balance of terror, right?

So I don't want to minimize Thomas Schelling's achievement.

It was great.

But it was like a two or three-dimensional problem, right?

I think with AI you have more like a – more than three-dimensional problem.

So it is consequently more difficult to figure out how do you create a game theory solution to this that is kind of stable, right?

I think what you can say at a minimum right now is there are ways in which the US-China relationship today –

resembles that of the United States and the Soviet Union in the 1950s, i.e.

before the Cuban Missile Crisis, right?

And I don't want to exaggerate this too much, but the sort of diminution in exchange in communication over the last six or seven years is really alarming.

So give a simple example on this.

When Trump came into office the first time in 2017, he inherited structure where there were approximately 100 –

working level dialogues between US government and Chinese government officials, plus all of this investment in private sector stuff that was going on.

But just at the formal level, there were over 100 dialogues.