-
agi timeline
2026-01-22
Human-level AI will come, but not built on LLMs; needs major conceptual breakthroughs, not in a year or two.
«We'll have AI with human-level intelligence, but not built on LLMs.»
-
other
2026-01-03
↺ 입장 변경
On leaving Meta: insists on his scientific autonomy.
«You certainly don't tell a researcher like me what to do.»
-
agi timeline
2026-01-03
↺ 입장 변경
Sharply states LLMs are a dead end for superintelligence; leaves Meta to found AMI Labs.
«LLMs basically are a dead end when it comes to superintelligence.»
-
other
2025-01-23
Working on world models: systems that build a mental model of the world, not just manipulate language.
«LLMs are good at manipulating language, but not at thinking... systems that build mental models of the world.»
-
agi timeline
2025-01-23
Says the current LLM paradigm has a short shelf life of 3-5 years before a new architecture.
«The shelf life of the current LLM paradigm is fairly short, probably three to five years.»
-
jobs
2024-10-23
AI will amplify human intelligence rather than replace people; it should empower, not threaten.
«AI assistants will eventually be smarter than us, but we should feel empowered, not threatened.»
-
safety
2024-10-12
Calls fears of AI's existential threat complete nonsense.
«You're going to have to pardon my French, but that's complete B.S.»
-
safety
2024-05-18
Worrying about controlling superintelligence is premature; there's no design even smarter than a house cat.
«we need the beginning of a hint of a design for a system smarter than a house cat»
-
regulation
2024-01-26
Rejects calls to urgently regulate LLMs as dangerous, calling that untrue.
«we need to regulate LLMs because they're gonna be so dangerous. That is just not true.»
-
open source
2024-01-26
AI's future must be open source for cultural diversity and democracy.
«The future has to be open source, for cultural diversity, democracy, diversity.»
-
agi timeline
2024-01-26
Argues LLMs are not a road to AGI and dislikes the term itself.
«They're not a road towards what people call 'AGI.' I hate the term.»
-
open source
2023-10-30
Fears a future where AI is controlled by a few firms; open source is therefore critical.
«This is what keeps me up at night»
-
agi timeline
2023-06-14
Predicts LLMs are a passing phase, to be largely abandoned for more robust algorithms.
«[LLMs] will be almost completely abandoned in favor of better, more robust algorithms»
