AI accidentally do a thing deleting millions or billions of dollars worth of information? AI literally killing people?? It’s just learning, sweaty
A human cashier not smiling at a customer who could quite literally be described in terms interchangeable with a description for a literal troll? goodness gravy, that’s no good, they deserve to be fired and die in the streets!
you might think “this is a badpost sir” but I never voted for badposting and I’ll never post there badly or otherwise
That’s a huge amount of the reason for AI being used, to obfuscate blame, since it’s really the fault of the people who decided AI would be able to make those decisions, but corporations and seemingly states are making an effort to prevent it from ever being interrogated that far.
This isn’t a badpost.
AI doesn’t have to go through 17 rounds of job interviews to put tens to hundreds of critical bugs into windows. They just let it do that.
they ask it not to do it and then it keeps doing it!
This isn’t a badpost, this is kind of insightful, I never really thought of it that way before, but you’re right, if I knew a person who was constantly just making shit up the way AI does, I’d never listen to them about anything, and is the main reason I don’t use AI for anything, but people are weirdly tolerant of the “mistakes” AI makes. Maybe it is the asskissing apologetics, but I also hate AI because it does that so much too.
So I guess the target audience for this stuff is something who demands their ass be kissed nonstop and only cares about being “right” rather than actually being “correct” about things. The same sort of person who will try to get a cashier fired for not sufficiently bootlicking and apologising loves AI because AI always folds and always tells them how right they are about everything, AI being wrong just gives them a chance to demand an apology from the AI, so they probably see that as a benefit, not a negative.
This is a good post and kind of the reason I come to this site. Great way of looking at it I’ve never seen expressed elsewhere and yet… It’s so obvious.
Good job.
Thank you this would be my job if I could get paid for it
Seconded o7
Developers: Make a mistake in production code
Managers: WHAT THE HELL IS THIS!? UNACCEPTABLE! FIX THIS IMMEDIATELY!
Clanker: Writes dogshit slop code that doesn’t work, it sucks ass, etc
Managers: Well you see, it’s still learning. It will GET BETTER. It had insufficient training data! We just need to adjust our prompts.
Write shitty code and pretend you vibe coded it.
Get promoted.
That’s what we call an art of the deal
There’s an episode of this website’s podcast
about the responsibility-resolving nature of tech in general. AI is an iteration on that. Actual people made all of the decisions, in reality, including the decision to have a hallucination machine go ham on whatever the topic happens to be, but of course as you note people focus instead on the machine itself and forgive it easily as if it is just an objective fact of reality of an oopsie that will be corrected.Replacing a human is far easier than replacing an AI model because there aren’t as many of the latter. They afford similar levels of “smol bean whoopsy daisy” good faith to CEOs, and for similar reasons.
