> ... the AI[Salami] was doubling down on the lie by mentioning reputable media outlets in an attempt to make the story more believable...
#ChatGPT was outputting fake links to articles that have never existed - why?
https://www.theregister.com/2023/03/02/chatgpt_considered_harmful/
#AlexanderHanff on dangers of relying on #AiSalami:
"Systematic Approaches to Learning Algorithms and Machine Inferences."

@taosecurity
I got little support elsewhere when I expressed concern. People I respect felt that it is "just" an enhanced spell checker, and attributing anything like a motive was naive.

The argument is that most biographical writing ends with an account of the death of the subject, and GPT3 is merely trained to make a pile of words in the correct order.

#ChatGPT #AlexanderHanff #FakeObituary

/Cont