Researchers say GPT 4.1, Claude 3.7 Sonnet, Gemini 2.5 Pro, and Grok 3 can reproduce long excerpts from books they were trained on when strategically prompted (Alex Reisner/The Atlantic)

Alex Reisner / The Atlantic:
Researchers say GPT 4.1, Claude 3.7 Sonnet, Gemini 2.5 Pro, and Grok 3 can reproduce long excerpts from books they were trained on when strategically prompted  —  On tuesday, researchers at Stanford and Yale revealed something that AI companies would prefer to keep hidden.

from Techmeme https://ift.tt/Z4xSvF1