Remember folks, ChatGPT lies with total conviction about easily verified facts, like this claim that solar eclipses can't happen anywhere in our solar system.
@cstross the analogy i use is AI is like asking Boris Johnson a question. You are going to get a clever sounding answer, with the apparent signifiers of knowledge, delivered confidently. However on investigation it may turn out to be completely fabricated.
@peterainbow@Inquisitive I saw him on a bicycle from the top deck of a London bus once. Asshole was ignoring the rules of the road and inconveniencing dozens of other people (he forced the driver to brake hard to avoid turning him into roadkill).
@cstross It's actually true, as stated. Gemini only stated that total planetary eclipses cannot happen, which is true. We only get isolated dark areas on a planet as its moon passes in front of the sun. It may not be the answer that the questioner asked, but hey... humans do that too. ;-)
@Npars01 All this is absolutely true! But NONE of this truthiness will get through to folks whose only interaction with it is when Microsoft Copilot pops up to offer them help, or the first page of Google search contains nothing but results that start with "Certainly! Here is ..."
@cstross Slight nitpick: It doesn't 'lie' so much as generate text that contains false information. There's no interiority or intentionality to it, it's just the most probable continuation (weighted by a preference model)
@pettter@cstross A liar is not immoral, they're amoral. A liar is someone who will say whatever yields the greatest gain for them, unbothered by whether the statement is true. That simply doesn't enter the liar's calculus. A liar is empathically NOT someone who's every single statement is a lie (that's absurd) and certainly not someone who tells untruth because it's fun to tell untruths. A liar is simply uncerned with the real state of the universe and says whatever gives them the greatest gain.
Which is 𝑒𝑥𝑎𝑐𝑡𝑙𝑦 what AI is doing: output whatever maximizes your reward function, unbothered by factual veracity.
@pettter
I'd argue for #bullshit in the Frankfurtian sense of it not caring about truth, just about outcomes - but it doesn't actually care about anything because, yeah, it's not capable of caring. @cstross
@FeralRobots@pettter If you are trying to explain "AI" to ordinary folks they will zone out the instant you use any word longer than "lying". So don't do that. Frame it in terms they can understand. If they ask for more details, then supply them. OK?
We have a HUGE mass education problem ahead of us. Don't let's make it harder by overestimating the level we've got to work at.
It's similar to anthropogenic climate change. Which HAS TO be shorted to "global warming" in the vernacular or a chunk of the public just won't get it.
And you're up against BP, Exxon, Shell, Aramco &c's propaganda.
You can be technically correct and still lose the argument because most people can't handle abstract factual reasoning. And we can't afford to lose these arguments.
@cstross@pettter
Sure, but I think we're mostly shitposting on this. (FWIW, though, I think there's an argument for 'bullshit' vs 'lying.' It's more dismissive - I'd map 'bullshit' to contempt, & 'lying' to anger. Contempt is often more damaging than anger.)
@FeralRobots@pettter "Bullshit" is a culture-specific colloquialism that may not have the same connotations to all English speakers, especially where it's not their first language. "Lies" are more widely understood. And I ALWAYS have to consider that stuff I say on social media might go viral, so I try to make it count.
Add comment