I'd like to comment specifically on the conception of betting on AI 'achievements' (I think Marcus' bet is underspecified and kind of vague in all 5 of its points).

People shouldn't be betting on benchmarks because benchmarks can be and usually are gamed (see Goodhart's law). Also, most people couldn't give less f*ck if an AI can write an award-worthy poem (I personally don't care about any form of AI "art", any sort of text an AI can produce or really any meaningless "feat" it (as in the general category) becomes capable of). The only worthy bets are ones that discuss economic impact. How many people will be structurally unemployed because of AI by year X? Will it lower or increase the GDP growth rate and by how much? Will it shift the balance between labor and capital and how? Etc.

So more meaningful bets and less benchmark bullshit that doesn't matter, please.