The a16z Show cover image

Marc Andreessen on AI, Technology, and the Future of Humanity

The a16z Show

00:00

Define Better Reward Functions

  • Design reward functions deliberately: avoid single-objective metrics like maximal engagement to prevent sycophancy or addiction.
  • Marc Andreessen recommends offsetting engagement rewards with safety, truthfulness and life-balance signals during post-training.
Play episode from 26:58
Transcript

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!
App store bannerPlay store banner
Get the app