News for hackers. Stuff that matters.
Re:ChatGPT: The Slickest Con Artist of All Time by rvz on Friday February 03, 2023 @11:26AM (#34643304)
Well an AI that can't explain itself transparently makes it untrustworthy for more serious use and applications since ChatGPT it is being celebrated as a great sophist. There is nothing new in this, other than 'train more data on the internet'.

The hype in ChatGPT reminds me of the short-lived hype of Clubhouse; the end result is the same: The bigger players will come in to surpass it, just like they did with GPT-3.

Re:ChatGPT: The Slickest Con Artist of All Time by vehemenz on Friday February 03, 2023 @04:49PM (#34648204)
I wonder. If the bigger players can make an AI good enough to be indistinguishable from human (but not quite), then wouldn't the AI slowly and incestuously train itself to become dumber, with nothing but AI-generated content left to "feed it?"