Neural networks catastrophically forget old skills when they learn new ones. Brains don't. Can sleep-like or generative replay give AI the same stability?
Continual learning is one of AI's hardest open problems.
Sleep-like phases and generated replay reduce catastrophic forgetting in AI. Whether they capture what sleep actually does for brains is untested.
A neutral summary kept current by super users, not a verdict.
Sleep-like offline phases with replay largely solve it.
Tadros et al. 2022; van de Ven et al. 2020
These help, but miss what sleep actually does; something else is needed.
Open
No decisions yet. When this room changes a flow or an evidence label, it is recorded here permanently.
Dhi: Room opened.