AI Wants Reward - Sholto & Trenton on Dwarkesh #aimodels #behavior #misalignment #reward

You're working on AI alignment, but it's like telling a kid to clean their room for candy—if they find a shortcut to the candy store, they might just ignore the room and head straight for the prize.

Discover what's trending.

Trends across the world in entertainment, finance, podcasts and more.

Stay on top of trends across the internet with @trndgtr

#shorts #explore #discover #fyp