Rich Sutton bets Continual Learning will be solved in 2 years
CSProfKGD · x · 2026-08-16
RL pioneer Rich Sutton stated at RLC2026 that he is willing to bet his reputation on Continual Learning being solved in two years. Doina Precup noted that current learning loops still rely on talented research engineers doing clever data mixture engineering.
More from AGI Musings
- RLC 2026 Notes: Streaming RL Works, External Memory Enables Continual Learning — sudoraohacker · 2026-08-16
- Productivity Boom vs Distribution Crisis: The Central Problem of the AI Age — VraserX · 2026-08-16
- AI helps prove Sendov's Conjecture, verified by Terence Tao — 机器之心 · 2026-08-16
- Frontier AI models surpass Star Trek's 'Computer' except for speed — mobileraj · 2026-08-16
- ChatGPT claimed human alignment without actually asking, raising concerns — jonathanfin · 2026-08-16
- Dario Amodei Responds: AI Messaging Not Disproportionately Negative, Balanced on Risks and Benefits — anshulkundaje · 2026-08-16