3 comments

[ 5.0 ms ] story [ 16.6 ms ] thread
This Formula 1 season, I'm putting frontier models to the test - can they use their research capabilities to accurately predict outcomes of races?
But can it predict Ferrari’s Next Year (tm)?
Well, ~1 hour after the race: it got the bets correct. But it didn't expect that Piastri would crash on the way to the grid (of course, we haven't invented a future-seeing AI yet ;) ).

But this is very voodoo, it's asking 3 black boxes "What do you think", they answer, but you probably can't see their reasoning. Or can you?