I attempted one thing totally different at my buddies’ annual Oscars occasion this yr: I outsourced my poll to Anthropic‘s Claude.
It beat across the bush at first, so I requested it to only give me the winners. Claude obliged, delivering assured picks throughout practically each class. And it labored.
I gained the pool and walked away with a field of candies and bragging rights.
Alistair Barr/Enterprise Insider
However here is the twist: Claude did not even full the task absolutely.
It failed to choose a winner for Casting, a brand new Oscar class this yr. Perhaps the mannequin did not register the addition?
And in a few different award classes, Claude picked candidates who weren’t even on these shortlists. Take a look at the photograph above to see the place it went flawed.
Even so, my Claude-powered poll nonetheless outperformed everybody else’s on the occasion.
That appears like a glimpse of the jagged edge of AI: programs which are clearly highly effective, sometimes brittle, and nonetheless adequate to win.
Subsequent yr, I doubt I will be the one one displaying up with an AI-generated poll.
Join BI’s Tech Memo e-newsletter here. Attain out to me by way of electronic mail at abarr@businessinsider.com.
