Summary
According to a post on X by @ibab, Igor Babuschkin argues that pretraining turns humanity's freely shared writing and open-source code into models that companies then sell for huge sums. He does not think a strong copyright claim works, because models learn and synthesize much as people do. Instead, he proposes that the resulting models be released openly, and he says that after years of building LLMs he believes this will become the majority view.
AI-assisted summary based on the listed source.
What happened
Babuschkin wrote that pretraining distills humanity's freely shared writing and open-source code into models that companies then sell back for huge sums. He doesn't think a strong copyright claim works, since models learn and synthesize as people do, and proposes instead that resulting models be released openly,...
Why it matters
For companies that sell LLM access, and for developers who rely on those products, the proposal raises a question about who should benefit from models built on publicly shared writing and code. Open-source contributors and content creators may also be affected if the argument gains support. The main thing to watch is whether other AI researchers or companies take up the position, since this is so far a personal argument rather than a policy or company decision.
Signal Intelligence
Signal Strength 93%
Technical label WATCH
Public Interest 32
Category OPEN SOURCE
Reader Depth TECHNICAL
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
Public Interest components
Recognizable Entity Score 0
Practical Impact Score 26
Novelty Interest Score 94
Consequence Score 18
Curiosity Score 0
Shareability Score 50
Why this is here
VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).