The Decoder is a media outlet that reports on AI research, policy, and products. Recent coverage includes a two-year university study finding bans on AI hurt student performance, research on distinct internal reasoning patterns in models, OpenAI asking Congress whether an industry slowdown is legal, analysis of Claude Fable 5.1’s stylistic shift, and Hugging Face’s launch of ML Intern.
OpenAI Has Contractors Reviewing ChatGPT ChatsReporting says OpenAI has hundreds of contract workers reading and rating anonymized ChatGPT conversations (the 'Improve the model for everyone' setting is on by default), and users must disable it to avoid human review, raising privacy and data-handling concerns.
Two-Year Study Finds Banning AI Hurts StudentsA two-year university study compared banning AI, unguided use, and structured training; students in the no-AI group performed worst both years, and the researcher said their prior assumption was wrong.
Distinct internal patterns for reasoning stepsA new study finds that models exhibit distinct internal patterns for different written reasoning steps—calculation, formula retrieval, and deduction—with these patterns strongest in middle layers. This shows models internally process more than their visible chain-of-thought, affecting interpretability and AI safety analysis.
OpenAI Questions Legality of an Industry SlowdownThe Decoder reports OpenAI has asked Congress whether an industry-wide slowdown in AI development would be legal, seeking legal clarity on coordinated pacing of AI progress.
Analysis: Claude Fable 5.1 writes more plainlyArena.ai analyzed tens of thousands of benchmark responses and found Claude Fable 5.1 writes more matter-of-factly but also more verbosely than Fable 5, indicating subtle stylistic shifts.
Hugging Face launches ML Intern to run ML experiments via chatHugging Face launched ML Intern, a chatbot-integrated assistant that lets non-experts run machine learning experiments via conversation, lowering the barrier to ML experimentation.
The A.I. Daily