Ask to be corrected, and print the corrections with their sources

The article on generative AI carries a section of corrections (machine learning learns too; no stored graph inside a model; the prompt conditions; next-token prediction plus training on feedback; emergence) because Dinis asked for them: the book does the same wherever his picture and the evidence differ.

Decided by. Dinis, in Generative AI, not AI

Where it comes from

Generative AI, not AI Where I was corrected (article)
My memory, checked publish only what the record supports (newsroom)

What it answers in the 2018 book

Facts answered outline, Concepts
Being criticized is an privilege answered written, Life Patterns

Where it goes

Chapter Facts, evidence and the record, part Direction and judgement (drafting); topic Facts, evidence and the record.

Other ideas proposed for the same chapter: Make fact-based decisions: think out loud and let the record check you, Question everything: a claim earns its status by being refutable, Questioning, and asking for the evidence, is one of the most important skills now, They care about privacy; Snapchat's best feature was that images disappear, Following a rabbit hole is now a prompt away; build the agents and tools that find where things come from, Provenance and explainability are how you disagree rationally, Managing information so that decisions can be made is what work is, What cost hundreds of thousands of pounds is now a brief and a feedback loop, Be the best-prepared person in the room: briefings, pre-briefings, follow-ups, Provenance is a graph: where a claim came from is a chain of edges back to its source, A graph is trustworthy not because it is right but because it is where users, experts and tests can correct it, Stale pages become false memories, so a memory must say where each thing came from and name its own errors.

Node idea:asked-to-be-corrected in graph.json; see it in the graph. Source: admin/content/graph/ideas.json.