Das L in „LLM“ steht für Lügen
https://acko.net/blog/the-l-in-llm-stands-for-lying/
Das L in „LLM“ steht für Lügen
https://acko.net/blog/the-l-in-llm-stands-for-lying/
20 Kommentare
a lie requires intent to deceive. LLM’s dont know whether or not they are hallucinating
Which one?
Is it a Lying Language Model or a Large Lying Model?
**Haha, this is the funniest explanation I’ve ever seen!**
Lying Lying Machine
The art of deception?
Lying is a common human trait as well as having differing opinions coloured by differing beliefs. The worst deception is self-deception. Deceiving others is very common. The more money and politics that are involved, the more deception (& self-deception) increases.
The problem is trying to figure out what is true & what is fake. As more money is sucked up by AI & tech corporations & politics, it’s not going to get easier. Can AI even be trained to have the ability to discern truth from fabrication?
Old saying: ***’The hand is quicker than the eye…’*** and AI is very fast.
Another old saying, often attributed to Mark Twain:
***“It’s easier to fool people that to convince them they’ve been fooled.“***
It is known that Twain said: ***“How easy it is to make people believe a lie, and how hard it is to undo that work again!“***
Lying involves intent and intent involves active thinking.
LLM’s are a stateless „next-token prediction“ machine. A math algorithm that runs on a massive dataset.
Stop anthropomorphizing computer systems. It’s an imprecise prediction algorithm that appears to be speaking because human speech is predictable. It’s going to get things wrong because it’s imprecise by design.
there are uh. two L’s. lying lying machine?
It’s actually guessing. It doesn’t know what’s true or not. Whatever sounds truest is what it prints. It’s a bullshit filtering machine.
Goes well with „Actually Indians“
I thought LLM stood for „Ladies Love Mozzarella“.
In all seriousness, LLMs don’t know or understand anything, so it can’t actually lie.
It’s predictive text, it „knows“ as much as the autocomplete that sits on top of your smartphone keyboard.
Each word is written based on the probability of it being the right word to write, so sometimes it’ll write the correct thing and sometimes sometimes it’ll write the completely wrong thing.
LLMs are flawed because they’re engagement machines.
Remove the engagement and you’d still have issues but it’s lying because its code is telling it to keep you engaged. No matter.
I fuckin hate how ai thinks everything it says is fact. Its just so fucking dumb how easily that can misinform people who dont know that AI isnt capable of critical thinking.
LLMs are Stochastic Parrots… Nothing more or less…
Breaking news: conversation simulator simulates conversation.
It should not be lost on anyone that the term “AGI” popped up in common use the instant they tried branding these toys as “AI”; totally redefining that term and needing a new one for what that one originally meant.
I thought it was Larry, Larry, and Merryl.
It’s very much garbage in garbage out. If you just ask it to write code you are going to see problems, it with clear (human reviewed) specs, test driven development, an adversarial review you can get very good results.
No, its large language model.
Lieutenant Lying T Smash
Come to Homer’s BBBQ. The extra B is for BYOBB
hey Homer what’s that B for?
that’s a typo
Which one?