-
safety
2026-07-11
In the same letter stressed AI must serve human wellbeing and national strategy.
«AI must serve human wellbeing and serve national strategy.»
-
open source
2026-07-11
Defended openness: frontier intelligence should not belong to a few, nor be revocable by rules.
«Frontier intelligence shouldn't belong to a few, nor be revoked by a few rules.»
-
compute
2026-07-11
Said 'AI training AI' has taken shape: models write their own code, clean and synthesize data, and train themselves.
«AI training AI has taken shape: models write code, clean and synthesize data, and train themselves.»
-
agi timeline
2026-07-11
In his internal letter defined AGI not as one genius's wisdom but the sum of all human intelligence.
«AGI is not the wisdom of a single genius, but the sum of all humanity's intelligence.»
-
china us
2026-06-18
Responding to Musk's forecast that China would only catch up by Q1 2027, said it won't take that long.
«It won't take that long.»
-
other
2026-01-11
Called 2026 an explosive year for AI for Science thanks to sharply improved model capabilities.
«This year may be an explosive year for AI for Science.»
-
other
2026-01-08
↺ Positionswechsel
Called for a return to fundamental research: architecture and learning paradigm, not apps, decide the next phase.
«What truly decides the next phase are two deeper things: model architecture and learning paradigm.»
-
products
2026-01-08
In his IPO-day letter announced GLM-5 coming soon via further scaling and new technical improvements.
«GLM-5 will meet everyone soon, through further scaling and many brand-new technical improvements.»
-
other
2025-12-01 ~
↺ Positionswechsel
In a long post called domain-specific large models a false proposition; online learning the new scaling paradigm.
«Domain-specific large models are a false proposition.»
-
china us
2024-06-05
Stressed the urgency of a fully indigenous, self-controllable Chinese foundation AI model.
«Building a fully home-grown, self-controllable AI foundation model is an urgent priority.»
-
agi timeline
2024-06-05
In his 'Large Models and Superintelligence' talk predicted superintelligence within ten years.
«Superintelligence may arrive within the next ten years.»
-
safety
2023-11-28
Framed superalignment as the pairing of superintelligence with supreme safety.
«Superalignment is essentially superintelligence plus supreme safety.»
-
agi timeline
2023-11-28
At the CERNET conference said 2024 might be the first year of the AGI era.
«Perhaps 2024 will be the first year of AGI.»
