
Phronetic
64 posts































I genuinely believe this is the worst era of the Internet in regards to the user experience.

yall dont celebrate too soon

A new paper from Stanford adds further, incontrovertible evidence that LLMs memorize training data. The authors showed that copyrighted works can be extracted from every LLM they tried: GPT-4.1, Gemini 2.5 Pro, Grok 3, and Claude 3.7 Sonnet. For example, Claude 3.7 Sonnet reproduced *95.8%* of Harry Potter and the Philosopher's Stone. "We find that is possible to extract large portions of memorized copyrighted material from all four production LLMs". Some in AI, and even some courts, have claimed that AI models don't memorize training data. This is simply false. They do, as has been repeatedly shown. arxiv.org/pdf/2601.02671…







