I used Anthropic's Claude to pick Oscar winners at a party. It made odd mistakes, but still beat everyone else.

At this year's Oscar celebration with friends, I decided to take a unique approach: I let Anthropic's Claude AI predict the winners for me. Initially, Claude was a bit hesitant, but when I prompted it for direct answers, it confidently provided its selections for nearly every category. To my surprise, this AI-assisted strategy paid off. I ended up winning the pool, earning both a box of chocolates and the coveted bragging rights. However, there was an interesting twist—Claude didn’t complete the task entirely. It overlooked the new Casting category introduced this year, perhaps failing to recognize this addition. Furthermore, in some categories, Claude selected nominees not even listed on the official shortlists. Despite these errors, my AI-generated ballot still outperformed everyone else's predictions at the party. This experience offered a fascinating glimpse into the complexities of AI technology: remarkably capable yet occasionally flawed. I have a feeling that next year, I won’t be the only one relying on AI for Oscar predictions.