Variation and Continual Learning

In working with cultural evolution simulations with LLMs, I’ve noticed it can be difficult to get enough variation for selection to have much bite. A common view is that continual learning may be the final missing piece before we get AGI. But this might also solve the variation problem: agents with long—and, importantly, different—life histories will be far more varied. If so, continual learning might also unlock powerful cumulative cultural evolution in AI populations.