AI hallucinations

cognitive science ai brain not database a jellyfish with a domed bell and trailing tentacles

AI Behaves More Like a Brain Than a Database — Cognitive Science’s Role in Its Origin Story Helps Explain Why

Two cognitive scientists argue that modern AI behaves more like a brain than a database, and that its origin story explains why. We trace the psychology behind neural networks from Hebb and Rosenblatt’s perceptron to backpropagation, check the primary sources, show how research on human memory explains AI hallucinations, inconsistent answers and sycophancy, and set out practical rules for using these systems at work.

Read more
ai summaries misleading distort human memory a yield sign on a post

Misleading AI-Generated Summaries Can Distort Human Memory

A Georgetown and University of Washington study found that a ChatGPT summary with one wrong detail cut correct recall of a traffic sign from 83.6% to 44.8%, whether people thought AI or a human wrote it. Every AI summary the team generated also omitted central facts. We explain the method, the numbers, the 50-year-old memory research behind them, and the controls organisations need.

Read more
lawyer fined 5k ai hallucinated witnesses murder case a balance scale two flat round pans level beam

Lawyer Fined $5K Over AI-Hallucinated Witnesses in a Murder Case

The New Mexico Supreme Court held Santa Fe defence lawyer Stephen Aarons in contempt, fined him $5,000 and removed him from a life-sentence murder appeal after his ChatGPT-assisted brief quoted four witnesses who do not exist and three who never said what it claimed. We read the 9 September order as quoted by the outlets that saw it, the 21 August hearing coverage, the trial record behind the appeal, and the 2,039-case AI hallucination database, and counted where this case departs from the fake-citation cases that came before it.

Read more
CHAT