Man, I love when I tentatively explain a concept to GPT to see if I've understood it right, and it tells me "Exactly! You've got it right".
Really builds my confidence.
Why are any oxygen-breathing teachers still employed?
Man, I love when I tentatively explain a concept to GPT to see if I've understood it right, and it tells me "Exactly!
Based, who cares about learning factually correct knowledge anyway?
>language models hallucinate a small percentage of the time in certain contexts, but I'm going to pretend it's 90% because performative cynicism makes me feel elite
everyone you know secretly thinks you're really boring
>anally flustered
uh oh..
>a small percentage
Anon, ask it some math questions beyond basic arithmetic, or physics questions. It often gets answers right but the reasoning totally wrong, or somehow gets the answer of 1 or 0 for the derivative of a function.
Yes, language models aren't calculators and you're better off using Wolfram Alpha for that sort of thing. You're not redpilling anyone on language models by pointing this out. It doesn't mean they're useless.
You're the one who used the word useless. OP was acting like you can trust everything these models say which is demonstrably false.
This is simply not true, they're pretty good at advanced math tbh
T. Math postgrad
The other day I asked it to revise a sentence for me and it gave me an answer confidently, but it felt awkward to me for some reason. I told GPT that the sentence felt "clunky and awkward, but maybe it's just me? Give me your honest input please" and it said "I understand your feeling completely, let me try formulate it in another way" and it gave me an epic sentence. I could almost cry
Most intelligent GPT user
You know it says that pretty much no matter what.
>hey, how do you do X in Y environment? is it Z?
>yes, the way to do X in Y environment is Z
>it isn't Z.
That pretty much sums up my experience with ChatGPT. 1/10 "revolution". Would not purchase.
Why am I getting the stupid verification "point picture in right direction" bullshit whenever I enter a prompt? Is it because I am on Linux and globohomo doesn't like that?
No one likes Linux.
i ask it to explain things like warner herzog and i turn it into an audio file using his voice.
https://vocaroo.com/1dXaZS2Y03G0
AI gays are worse than cryptogays.
>cancel the future, let's just have the status quo forever
no
What happened to the blockchain being the future?
>asking satan for knowledge
You will never posses a real intellect
>Exactly! Got it right!
Take answers like this with a grain of salt. If it ever begins with a hard yes or no, it's next output is going to be trying to justify it's first word. Ask it to analyze the situation first, explain it's reasoning, and only then give a final answer based on its analysis.
You just have to know the quirks of how LLMs operate and be diligent. Humans get shit wrong all the time as well.
I get suspicious that ChatGPT is being sycophantic when it tells me I'm correct too often.
Sometimes I do the opposite of what OP says and state the opposite of the conclusion I arrived at. If ChatGPT still agrees I start putting less weight in its answers
ChatGPT is going to revolutionize education for the better
kids who don't care about learning X will be able to effortlessly cheat their way through
kids who do care about learning X will have a patient mentor willing to answer all of their questions
everyone wins except the entrenched education system
this
anti ai gays are too short sighted to see how useful it is to have a redditor answering your questions for free, and fast. dont tell me people dont add "reddit" to their search query to improve the results. you also wont have to deal with the stackoverflow "chads"