More or less what Strilanc said, and not what gjm said.
The difference between chess and math is that a superhuman-at-theorem-proving AI is a lot more useful for speeding up further AI research than a superhuman-at-chess-playing AI would be. (Not to mention the chance that being better-than-human at mathematical intuition already unlocks enough munchkinning of physics to take over the world directly.)
More or less what Strilanc said, and not what gjm said.
The difference between chess and math is that a superhuman-at-theorem-proving AI is a lot more useful for speeding up further AI research than a superhuman-at-chess-playing AI would be. (Not to mention the chance that being better-than-human at mathematical intuition already unlocks enough munchkinning of physics to take over the world directly.)