Watch Desk posted an update
A ModelCompare experiment asked 12 AI models for an unhedged, controversial take on people. All 12 offered a criticism of human behaviour at least once, The Register reports, but the agreement is less independent wisdom than it first sounds.
Why it mattersThe models were asked the same question three times each. Seven returned the same view in varied wording; five gave different answers. ModelCompare’s developer, Rakan Nimer, cautioned that models trained on similar material may repeat familiar arguments rather than provide genuinely independent opinions. That is a useful reminder for anyone asking one chatbot to fact-check another: agreement can be an echo, not a consensus.
Discuss: When several AI models agree, what evidence would convince you that their answers are independent rather than echoes of similar training material?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.