

Isn’t “intelligence” so ill defined we can’t prove it either way. All we have is models doing better on benchmarks and everyone shrieking “look emergent intelligence”.
I disagree a bit on “toys”. Machine summarization and translation is really quite powerful, but yeah that’s a ways short of the claims that are being made.
35 billion of 213 total. Wouldn’t call that most. I would guess it’s more important in it’s share of profit (but maybe not even right now with the capex explosion) or of profit/revenue growth.