“Jerry Twerk hypothesizes that continual learning has not yet been solved because it likely requires research at a scale of compute and model capability available only to a few well-funded labs focused on other priorities.”