Can some smart bros please explain de Finetti's theorem in a language that I, a retard, can understand? I've been trying to look it up, but I've been getting filtered by all the terms and context.
Alternatively, what would be a good place to start in a path to understanding all this?
It's cope for AI tards. I'm not even joking. They want their AGI and they want to believe that GPT-4 is the first real step towards that. Ultimately, it's still just an autocomplete.
It may not be AGI, but clearly LLMs can produce valuable results, and people who made them didn't just pull them out of their asses lol
If you are against LLMs at this point you are either not a programmer or never used it
It's like stack overflow on steroids
It's as if the biggest experts on stack overflow were waiting for you to ask your dumb question and immediately answer you no matter how stupid the question is and produce a reply exactly to your specifications
And that's just part of what it can do
It's a fucking miracle that's what it is
> Ultimately, it's still just an autocomplete.
So just like your brain then
I didn't know humans are turing machines, considering one is not limited in ways the latter is
Anon, I hope you put more thought into your... thoughts than just picking the next more probable tokens in a string.
I never do
Just blurt out whatever comes to my brain
Sometimes I hear a little voice saying "DON"T SAY THAT!!!!"
Then I make sure I repeatedly say it over and over again
Nobody tells me what to do
>why yes, I do have Tourette's, how could you tell?
Why don't you get Bing AI to explain it for you?
That's what I do when I'm lazy to read papers
Why have it explain it to you when you can have it take action based on the paper? Thinking is deprecated.
LLMs are just as vulnerable to bias and being wrong as we are, based on their training data.
You should treat them as just another person in the room. High in knowledge, low in wisdom
No dude they're a literal god, haven't you read the papers? AGI will realise tomorrow
Yeah, I read that it would cost $1000 to permanently cripple any dataset scavenging efforts from now until eternity
I'm sure we'll have an even cheaper way if we just ask chatGPT to plan that! Even then if someone asks it to fix the data our efforts will be in vein
Thank you, anon!
After getting Bing AI to explain de Finetti's theorem and its relation to AI models, I have concluded that the OP was word salad and not meant to prove anything other than a plea for ego stroking.
The theorem states nothing more than an abstract claim about some probabilistic models, and has a tenuous connection to AI models, making no claims about AGI in general.
It's sort of like saying, "Hitler was a vegetarian therefore he did nothing wrong."
While both may be true statements, there is no clear logic connecting one statement to the other.
It's a trivial observation that if you are "just predicting the next word" on a very difficult multi choice test that you have never seen before you must be doing something very smart.
>y'all Ö/