GPT-4 shifted opinions more than humans did, once it knew who it was arguing with
The finding comes from a single peer-reviewed trial, and it is the sharpest evidence anyone has produced for the argument that AI could learn to manipulate people at scale.
The paper credited with starting the LLM era never trained a language model
Vaswani et al.'s 2017 paper reported BLEU translation scores on two language pairs and never once trained the kind of model it is now credited with inventing.
GPT-4's 90th-percentile bar score traces to an uncited footnote
OpenAI's "90th percentile" rests on two sentences in a co-author's footnote that name no chart and no comparison group. Recomputed against attorneys who actually passed the exam, the same 298-point score lands near the 45th percentile, and the 15th on the essay questions closest to real legal work.
A transformer redraws a word's vector for every sentence it appears in
Word2vec's famous king-minus-man-plus-woman trick ran on a vector frozen once at training time, and most explanations of how today's language models represent a word still describe that same frozen kind.
Ofqual built a model that graded schools, then handed the grades to students
England's 2020 exam algorithm downgraded 39.1% of A-level grades by fitting each student to their school's 2017-2019 record.