Zero-experiment RLT paper claiming 'infinite reasoning depth' goes viral, draws skeptics

generativist · x · 2026-09-13

Yifan Zhang announced the Recurrent Looped Transformer (RLT), claiming it gives Transformers "infinite reasoning depth" and calling for pacing frontier progress until safe superintelligence is achieved.

Alexia J.M. and others pushed back: a paper without a single experiment is going viral — "either people have not looked at the paper or they asked an LLM to do it for them." The community widely views it as a hype-driven release.

Original post →

More from Fun

• The 130-140 IQ doom loop: why some smart people can't be optimistic — djcows · 2026-09-13
• Coinbase's very first hire was a former lumberjack — and he's now a billionaire — _AustinCalvert_ · 2026-09-13
• The SSI Name Joke: 'Safety First' Literally Written into the Company Name — umesh_ai · 2026-09-13
• The 'show your work' teachers were right all along, in the age of AI reasoning — lpachter · 2026-09-13
• The EA polycule is taking over ASI: insiders on safety's closed circles — TinfoilTricorn · 2026-09-13
• AI image flunks walking pose, and one dev had the perfect quip — flngr · 2026-09-13

Fun channel →