A model can't count the words it's writing
GPT-4 Turbo blew past its own word-count limit on 49.3% of prompts in one benchmark, because generation keeps no running count of length and can't take a word back.
GPT-4o clears 61% of τ-bench tasks once and under 25% of them eight times in a row
τ-bench scores an AI agent by whether a simulated customer's conversation ends in the correct database state, while eight straight successes collapse that score.
Generative models fed only their own output collapse within nine generations
A follow-up paper found that accumulating real data instead of replacing it avoided the collapse in every experiment the researchers ran.
The FCC's $6 million fine over the Biden robocall survived Steve Kramer's acquittal
Kramer's father paid a magician $150 to clone the president's voice for a robocall sent two days before New Hampshire's 2024 primary.
The report that coined ‘foundation models’ ran no experiment
A 114-author, 214-page report chose the name to signal architectural stability and safety, then admitted, in the same section, that it could not say whether the foundation was trustworthy.