RLHF Is Cr*P, It's a Paint Job on a Rusty Car: Geoffrey Hinton
a year ago
- #Critique
- #AI
- #RLHF
- Geoffrey Hinton, a pioneer in AI, criticizes RLHF (Reinforcement Learning from Human Feedback), calling it a 'pile of crap' and comparing it to a superficial 'paint job' on a flawed system.
- RLHF is a machine learning technique that integrates human feedback to refine AI behavior, particularly useful for complex tasks like natural language processing.
- Hinton argues that RLHF merely masks underlying issues (e.g., biases, inaccuracies) without solving fundamental problems in AI design.
- Hinton's critique reflects broader concerns in the AI community about the shaky foundation of current AI development approaches.
- Other experts, like Meta’s Yann LeCun, also doubt that current AI techniques will achieve human-like intelligence or sustain progress indefinitely.