Abstract
AI progress is accelerating, and now the leaders expect "artificial general intelligence" (AGI) in 2 to 3 years. While difficult to believe, we must prepare for the possibility. We can take some lessons from previous episodes in which computers surpassed humans, for example in chess. We survey current trends in AI for math and discuss conceptual studies of mathematical copilots and systems for autonomous mathematical discovery. These seem likely to become truly useful in the next few years.
