"It makes clear that LLMs will enable us to simulate goal-directed people, and watch as complex social and behavioural dynamics unfold."
Here's the sad problem with all of this: none of it is predictable. At best, it's probabilistic ... and that presumes our data and our agent generations are correct and proportionate, which is a ridiculous ask. Yes, you can Monte Carlo the crap out of this, but even the agents themselves are not deterministic. And that's the fly in the ointment. We all know what it's like to both desire ice cream and desire avoiding the effects of too much of this. None of this complexity can be captured yet.
"It makes clear that LLMs will enable us to simulate goal-directed people, and watch as complex social and behavioural dynamics unfold."
Here's the sad problem with all of this: none of it is predictable. At best, it's probabilistic ... and that presumes our data and our agent generations are correct and proportionate, which is a ridiculous ask. Yes, you can Monte Carlo the crap out of this, but even the agents themselves are not deterministic. And that's the fly in the ointment. We all know what it's like to both desire ice cream and desire avoiding the effects of too much of this. None of this complexity can be captured yet.