Natural abstractions don’t overlap and Most cognitive systems learn subsets of the same abstractions basically tell us that the set of all natural abstractions is rather small and that natural abstractions are discrete, they don’t overlap.
But do they overlap between cognitive agents? Is it likely that we learn the same abstractions as an alien? (Chan et al., 2023) argue that we likely do. If we don’t, every general purpose cognitive system is likely to be able to learn any abstraction. If we wouldn’t be able to learn any natural abstraction, an AI might learn some natural abstraction that we couldn’t understand.
Source: Chan, L., Lang, L. and Jenner, E. (no date) ‘Natural Abstractions: Key claims, Theorems, and Critiques’. Available at: https://www.alignmentforum.org/posts/gvzW46Z3BsaZsLc25/natural-abstractions-key-claims-theorems-and-critiques-1 (Accessed: 19 March 2023).