Revisiting December 2024 AI forecasts for math and cybersecurity
A post highlights researchers’ December 2024 aggregate forecast giving AI a 50% chance of solving a Millennium Prize problem around 2054—and argues that progress can make expert predictions look outdated.
TLDR
One post says AI researchers’ December 2024 aggregate forecasts put a 50% chance of AI solving a Millennium Prize problem around 2054, finding and patching a security flaw in a system with over 100,000 users in the early 2030s, and matching the best human Putnam competitors around 2030. A separate post argues that expert AI forecasters are good at predicting progress on model benchmarks but poor at predicting AI’s real-world impact and when it will arrive.
Combined views
119K
7 Sources, first seen ago
