pull down to refresh
80 sats \ 3 replies \ @optimism 11h \ on: The Most Insidious Trick Of AI Language Models AI
This is correct. AI is training AI. Search for LLM-as-a-judge in papers. Most have this, and therefore, these aren't trained by a human but by another LLM (often chatgpt) and therefore inherit biases.
Most have this, and therefore, these aren't trained by a human but by another LLM (often chatgpt) and therefore inherit biases.
Whooooow….. what you are implying is that the current AI is nothing more than turtles stacked higher on turtles! The AI problems are stacking themselves one-upon-another without circease. Isn’t it about time to train another AI from scratch to avoid previous problems? After all, these tech broz are hauling in billions upon billions of dollars to make their products. Sometimes people are even saying this is the new big thing!
reply
Yes. OpenAI employs a few thousand, not a few million, and afaik all the other LLM companies/divisions employ less. I agree that it's a bit of a mirage if you forget to take a healthy mountain of salt while listening to the guys that are raking in billions of fiat.
If everyone would take the salt, 3 things would happen:
- On average there'd be a lot more laughter, because these guys are clowns
- ChatGPT and Gemini would have significantly less users, i.e. no normies that got gaslit into using it
- Salt Lake City will have to rename to Ravine City because we'll run out of salt quickly with the rate these guys are yapping bs
reply
i just think that Scam Altman is really pulling a number on us with this AI scheme. The Chinese pulled of a new AI with much less time and money, they said. Clowns will be clowns, but the funny thing about it is that the normies are sucking it up and laughing it out through their noses, but taking it seriously! That is what I find seriously out of kilter.
reply