AI makes you think you’re a genius when you’re an idiot : techtakes

[–] the_q@lemmy.zip 25 points 2 weeks ago

I wonder if that's why you get those folks that defend it so strongly because the chat bot gave them validation through digital ass kissing.

[–] WatDabney@sopuli.xyz 17 points 2 weeks ago* (last edited 2 weeks ago) (1 children)

Effectively, what LLMs do is exactly the same thing that mentalists do - they wait for "prompts" to indicate your area(s) of interest, then feed you strings of words that are statistically likely to be well received.

Or in much simpler terms, and by design, they tell you what you want to hear.

[–] Seminar2250@awful.systems 20 points 2 weeks ago* (last edited 2 weeks ago) (1 children)

Baldur Bjarnason has a piece from July 2023 called The LLMentalist Effect: how chat-based Large Language Models replicate the mechanisms of a psychic’s con^[https://softwarecrisis.dev/letters/llmentalist/] that you might appreciate, if you haven't read it yet. :)

[–] WatDabney@sopuli.xyz 10 points 2 weeks ago

I did lift that basic concept from an article I read, and I would assume it was that one. Thanks for the link.

[–] Kolanaki@pawb.social 14 points 2 weeks ago (1 children)

In reality it just makes you stupid, but faster.

[–] swlabr@awful.systems 26 points 2 weeks ago (1 children)

obligatory:

[–] dgerard@awful.systems 18 points 2 weeks ago

I'M DOING 1000 CALCULATIONS PER SECOND AND THEY'RE ALL FP4

[–] ceenote@lemmy.world 10 points 2 weeks ago

Wtf I love AI now

[–] Angelevo@feddit.nl 4 points 2 weeks ago

We did not need 'AI' to get to this point.. People are freaking out over a mirror.

[+] CrayonDevourer@lemmy.world -15 points 2 weeks ago (5 children)

You can set them up so that they don't do that. It's just the companies running the chat bots set them up that way because that's what people expect now.

[–] Seminar2250@awful.systems 11 points 2 weeks ago

yeah the trick is to already know you're an idiot

[–] self@awful.systems 9 points 2 weeks ago (1 children)

Pay no attention to ChatGPT, we purposely trained him wrong... as a joke.

yeah ok

[–] Seminar2250@awful.systems 4 points 2 weeks ago

my face to your foot style

[–] BlueMonday1984@awful.systems 8 points 2 weeks ago

"you can set the sycophancy engines so they aren't sycophancy engines"

I'll take "Shit that's Impossible" for 500, Alex

[–] zbyte64@awful.systems 7 points 2 weeks ago

It's still telling you what you want to hear but with a different aesthetic. This will always be the case for systems that can't even perceive it's own lies.

TechTakes