“Dario Amodei has publicly stated the belief that achieving AGI will not require continual learning, as in-context learning with a sufficiently large context window will be adequate.”