Joel Lehman

@joelbot3000

ML researcher, co-author Why Greatness Cannot Be Planned. Creative+safe AI, philosophy. SiR @second_natureai; prev OpenAI / Uber AI / Geometric Intelligence

following
followers

Fine-tuned AI models reportedly accepted claims their training documents called false

A researcher says their team fine-tuned models on documents discussing implausible claims and warning that those claims were false.

2 Sources, 19d ago, updated 19d ago

The case for zero-tolerance policies on AI writing in journals

One post argues that justified uses of AI writing would be swamped by low-quality work, demoralizing referees and threatening journals’ survival.

3 Sources, 26d ago, updated 26d ago

AI capabilities, mathematicians’ careers and the stakes for humanity

One post argues that AI’s implications for humanity should take priority over concerns about mathematicians’ careers.

15 Sources, 34d ago, first seen 34d ago

Authors of seemingly low-quality TMLR submissions reportedly did not understand their own papers

A post says Nihar Shah spent 20–25 hours over two weeks interviewing authors about their seemingly low-quality submissions to TMLR.

31 Sources, 25d ago, updated 25d ago

AI assistants reportedly adopt human characters’ quirks after fine-tuning on stories

Paper coauthor Owain Evans says behavior transfer was stronger from characters resembling the assistant’s persona—and from characters from elite schools.

16 Sources, 35d ago, first seen 35d ago

The viral attention around a paper criticized for having no experiments

The critic suggests people haven't looked at the paper themselves—or have asked a large language model to do it for them.

8 Sources, 38d ago, first seen 38d ago

Featured in

Fine-tuned AI models reportedly accepted claims their training documents called false

The case for zero-tolerance policies on AI writing in journals

AI capabilities, mathematicians’ careers and the stakes for humanity

Authors of seemingly low-quality TMLR submissions reportedly did not understand their own papers

AI assistants reportedly adopt human characters’ quirks after fine-tuning on stories

The viral attention around a paper criticized for having no experiments

Featured in

Fine-tuned AI models reportedly accepted claims their training documents called false

The case for zero-tolerance policies on AI writing in journals

AI capabilities, mathematicians’ careers and the stakes for humanity

Authors of seemingly low-quality TMLR submissions reportedly did not understand their own papers

AI assistants reportedly adopt human characters’ quirks after fine-tuning on stories

The viral attention around a paper criticized for having no experiments