LeCun: Open Models Needn't Pursue Absolute Openness
ylecun · x · 2026-07-16
Yann LeCun shared a discussion regarding "open models":
- Pursuing "openness for all data" is unrealistic and may involve IP infringement.
- Releasing only the training framework is insufficient, as others may not be able to reproduce the training at the same scale.
- In practice, open models that can be inspected and fine-tuned are more valuable.
The core conclusion is that while these models might not be considered "completely open," this level of openness is far more practical and useful in reality.
More from AGI Musings
- Risk of being killed by a doomer mob higher than by AI, quips mark_k — mark_k · 2026-09-11
- Quantum physicist told his team 'I think AGI is here' — Forbes on what convinced him — Vailhem · 2026-09-11
- DeepMind alignment researcher signs open letter urging coordinated AI slowdown — vkrakovna · 2026-09-11
- WIRED: recursive self-improvement and rogue agent swarms spook AI researchers — nordicinst · 2026-09-11
- People Neglect Human Agency Both Ways: Exaggerated Doom and Complacent Optimism — jankulveit · 2026-09-11
- Garrison Lovely's 'Obsolete' on AI Replacing Labor Lands September 2026 with Heavyweight Blurbs — GarrisonLovely · 2026-09-11