After taking Berkeley AI course part 1 it feels like “getting stuck forever” is one of the standard game AI bugs. Makes you wonder if some if the mental mechanisms we have evolved to handle that.
But he certainly expected the “increment any counters” objective function would not map to perfect play, at least by the time he got to Tetris. It was more of a case of “well, this algorithm is probably not going to beat any of these games, but I’m not sure of all the ways that it will fail.”
I don’t understand why the algorithm didn’t pick up on his score in Tetris monotonically increasing—unless he was just such a bad player that the number of rows also monotonically increased?
It did pick up on his score increasing. But you get a few points just for putting a block on top of another block, and the search didn’t look far enough ahead (or wasn’t comprehensive enough) to spot that making lines would give even more points.
Also, the behaviour at the very end is very much “oh we didn’t expect that to happen forever when we wrote that utility function”
Wireheading :)
After taking Berkeley AI course part 1 it feels like “getting stuck forever” is one of the standard game AI bugs. Makes you wonder if some if the mental mechanisms we have evolved to handle that.
Boredom would seem like one obvious candidate.
It was published on April 1.
But he certainly expected the “increment any counters” objective function would not map to perfect play, at least by the time he got to Tetris. It was more of a case of “well, this algorithm is probably not going to beat any of these games, but I’m not sure of all the ways that it will fail.”
I don’t understand why the algorithm didn’t pick up on his score in Tetris monotonically increasing—unless he was just such a bad player that the number of rows also monotonically increased?
It did pick up on his score increasing. But you get a few points just for putting a block on top of another block, and the search didn’t look far enough ahead (or wasn’t comprehensive enough) to spot that making lines would give even more points.
I was referring to the “pause just before dying” behavior, which would have persisted even with enough search depth to make lines.
The finish line… is juusst around the bend. I’ll pause this game.… so our love will neeeevvver end...
I prefer the line where he implies eating a magic mushroom is Mario’s drugs.