Not sure a hugely topical and well-reported piece in The New York Times needs any amplification from me but….this just in … “the Times has uncovered nearly 50 cases of people having mental health crises during conversations with ChatGPT. Nine were hospitalized; three died”
You should read their report about what went on the inside, at OpenAI.
A big part of the culprit? Maximizing metrics for user engagement.
Lots of internal warnings were ignored.
Here’s Kashmir Hill’s own summary, followed by a gift link to the essay. It’s long but with lots of new insights into how OpenAI rolls — and by extension insights what that might mean for the future of AI safety.
You can read the essay here.