Tim Scarfe: belief

Transcript context

…. 1, what will happen if we take those types of models and put them in very new situations that are not based on just these puzzles, But for example, we're asking them to optimize a paperclip factory. Now that's a situation where we should be concerned how well will it do at actually inferring what we mean by what we say. And the second is data efficiency, which is how much data did it have to see to build this model. If it was a ton of data, then it's gonna be problematic if we have these new situations where we wanna teach them to model people's behaviors in this new place. If it requires a ridiculous amount of data, then it's always gonna sort of be slow to learn these things and always be at risk of not generalizing well when we put them in these new situations. So my answer here is nuanced, which is I think if by weak theory of mind, we mean solving puzzle questions, I think it's very hard to say that ChatGPT does not have some model of human behavior. But I do think the human and primate mechanism for doing so has a data efficiency advantage and a mechanistic synergy advantage. In other words, we can use ourselves to reason about things that is relevant. And if we wanna have these systems do a good job listening to human requests, we shouldn't translate performance on false belief tests to believing that they'll do a good job correctly inferring our intent to knowledge in new situations. Yeah. I I would agree with that. I I think chat GPT, it's in the the world of text, and it's learned all of this structured narratology and things on Reddit and things on Twitter. And as we were saying last week, you know, language has evolved to be very simple as a it's learnable by children. It has a small subspace. But it is it is a real kind of generalization over human behaviors, and it's and it's in this very low resolution substrate. Whereas in the Machiavellian apes example that we were talking about before, these are agents performing real time sensing and inferencing and making like, you know, in the moment judgments and they're they're in this continuous sensor domain where they have many, many different types of signals, you know, visual signals, sound signals, and also, you know, memory of of what happened in in those dynamics just before. So, it feels like a a difference in kind to me between those 2 situations. But it is remarkable that in the the GPT domain, any kind of theory of mind could could work. 1 good example of this, I think, is is there a difference in our human ability to predict behavior between a car and a person? So the brain is always able to model things it observes and simulate it and predict what it will do. So I can look at a car and I can imagine different colors of it, I can imagine what will happen if I drop it and it rolls down a hill. I can, we build models of things all the time. We build models of computers and models of, know, I'm just looking around the room of books. So the brain produces models of things. Is the way that the brain produces models of other human behaviors exactly the same, or is there some unique advantage? And my my argument is that there's something unique happening when I'm building a model of another person, which is I'm leveraging my own inner simulation of things as a as a useful prior to try and predict what other people will do. And so chat GBT models human behaviors in, to draw a crude analogy, the way we would model a random object, which is I'm only modeling it based on seeing its behaviors in certain situations with the data I received. On the other hand, when we model someone else's behavior, we're doing some form of projection and using the prior of how we would behave. And we probably bootstrap part of our model of human knowledge and intent based on our own introspection. And I think in that way, it is a difference in kind.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.