aifaq.wtf
"How do you know about all this AI stuff?"
I just read tweets, buddy.
#tweets
Page 17 of 37
"How do you know about all this AI stuff?"
I just read tweets, buddy.
Page 17 of 37
Welcome to the future! Story here (paywall skip)
It's honestly worth reading, every paragraph is a new list of Bad Stuff! Here's an example of doubling down on hallucinations:
The chatbot responded quickly, stating that Funiciello was alleged to have received money from a lobbying group financed by pharmaceutical companies in order to advocate for the legalization of cannabis products. But the entire corruption allegation against Funiciello was an AI hallucination. To “back up” its baseless allegations, the chatbot linked to five different websites including Funiciello’s own website, her Wikipedia page, a news article where the lawmaker highlights the problem of femicide in Switzerland, and an interview she gave with a mainstream Swiss broadcaster about the issue of consent.
And in its (ineffective) race to protect from the Bad Stuff, it isn't even good at the easy stuff:
In their study, the researchers concluded that a third of the answers given by Copilot contained factual errors and that the tool was “an unreliable source of information for voters.” In 31 percent of the smaller subset of recorded conversations, they found that Copilot offered inaccurate answers, some of which were made up entirely.
The guide can be found here
This isn't LLMs, but scoring systems for some reason fall under the banner of AI these days.
ChatGPT does better (or at least talks more) when you offer to tip it. Seems slightly better than the alternative of threatening murder or nuclear warfare to get things done. Now slighly concerned that we found this approach out like a year after the "if you don't answer I will throw children in a trash compactor" version.
I've posted this already but it's just so fun. This thread specifically has a million and one examples.
If we want to get historic, joge-e flip faces are a 19th century Japanese version of one of the visual tricks:
ChatGPT can reveal training data if asked to do specific dumb things, so now it's against the rules to do specific dumb things.
I want to reference rules lawyering even though it's absolutely not rules lawyering. But lawyering through rules, certainly.
