One thing I found fascinating about watching Claude play is it wouldn’t play around and experiment the way I’d expect a human to? It would stand still still trying to work out what to do next, move one square up, consider a long time, move one square down, and repeat. When I’d expect a human to immediately get bored and go as far as they could in all directions to see what was there and try interacting with everything. Maybe some cognitive analogue of boredom is useful for avoiding loops?
That is in fact a defect of these models and one of the things you can of want to scream into the screen after it, say, doesn’t walk more than 20 tiles to the right having spent days looking for an entrance that is 30 tiles to the right. Or when it doesn’t explore the bottom left of a room where the answer is because it’s convinced that’s not where it is.
Or the fact that it’s in week 2 of Silph co, driven by the fact that it’s convinced what it’s looking for is not an item on the ground, and not picking up any items even when walking right next to them, when in fact its goal is an item on the ground.
One thing I found fascinating about watching Claude play is it wouldn’t play around and experiment the way I’d expect a human to? It would stand still still trying to work out what to do next, move one square up, consider a long time, move one square down, and repeat. When I’d expect a human to immediately get bored and go as far as they could in all directions to see what was there and try interacting with everything. Maybe some cognitive analogue of boredom is useful for avoiding loops?
That is in fact a defect of these models and one of the things you can of want to scream into the screen after it, say, doesn’t walk more than 20 tiles to the right having spent days looking for an entrance that is 30 tiles to the right. Or when it doesn’t explore the bottom left of a room where the answer is because it’s convinced that’s not where it is.
Or the fact that it’s in week 2 of Silph co, driven by the fact that it’s convinced what it’s looking for is not an item on the ground, and not picking up any items even when walking right next to them, when in fact its goal is an item on the ground.