From the Business Insider article:
"There are safeguards in place to protect sensitive content, and the data is not used for any other purpose," a Meta spokesperson told Business Insider.
This is standard PR. There is now a sign at my local supermarket about the facial recognition software they have in place. "Don't worry," it begins. Supposedly the data on anyone who is not a known criminal, etc., is immediately destroyed. We are supposed to simply take their word for it, but given the creep of the usage of these things, it seems likely that as much usage is made of all these technologies as is profitable and convenient for the users. Excuses will be prepared in advance for whatever is discovered later, when the systems are well entrenched.
There is no particular reason to suppose that any current AI systems are sentient. But there is such a thing as AI psychosis, it seems. I expect people know of the case of Eugene Torres.
For an in-depth guide to how LLMs actually work, Melanie Mitchell's
Artificial Intelligence: A Guide for Thinking Humans, goes into some detail. There is nothing there to indicate sentience or anything other than what Emily Bender has called a "stochastic parrot". We are, of course, well beyond the Turing Test being passed, but the whole premise of the Turing Test is the question, Can a machine fool a person into believing it is intelligent? Or, a better way of phrasing that, Can a person be fooled into thinking a machine is intelligent? The answer, clearly, is yes. We can interact with machines (specifically LLMs) and have the sensation of communicating with something.
dflewisreviews.wordpress.com
