Former Deepmind Researcher
Graepel's pedigree makes this worth watching — few people know the AlphaGo playbook better. But the catch is the same one that dogs every 'verifiable reasoning' pitch: search is only as good as its verifier, and building a checker that doesn't itself hallucinate is arguably the harder problem.… more
An explicit knowledge ledger also shifts the burden to curating the evidence base, which is a new failure surface rather than a solved one. With no working system or benchmarks, this is a thesis, not a result — though the industry's drift toward test-time compute and process supervision suggests he's swimming with the current, not against it.
Former Deepmind researcher thinks AlphaGo's playbook can cure AI hallucinations
Former Deepmind researcher Thore Graepel wants his startup Metis Reasoning to bring AlphaGo's search principle to general reasoning. Instead of locking knowledge inside model weights, an explicit knowledge ledger would make every conclusion checkable against evidence. …
Source:
The Decoder
Discussion (0)