Don't want to start defending Leo, but he ain't wrong in this. If anything, this topic proves his point.
The numbers don't decide, the system is a lie. The river running dry, the wings of butterflies.
And you may pour us away like soup. Like we're pretty broken flowers.
We'll take back what is ours. One day at a time.
Uber wasted billions of dollars in tokens on AI spending. It got out of hand with idiotic employee tokenmaxxing, but still, prices will have to come down eventually or there is going to be a problem.
Don't want to start defending Leo, but he ain't wrong in this. If anything, this topic proves his point.
The point is, he has no point, no arguments are ever presented either.
Thinking he knows better than people who are scientists and actually work in this field speaks volumes of how disconnected from reality he is.
I very, very much doubt he even uses LLM's or has even a 1% understanding of the topic, for some reason he's just angry.
Computers has never been as fun and interesting as now due to LLMs, not even the internet was this fun when it was new. So I am very glad i can be happy and have an awesome time with it instead of being an angry tool.
I have no idea where this anger comes from either, if its just pure brainrot / MSM brainwash. It seems kind of similar to TDS to me, at least it's the closest thing i can think of, so something produced by (mostly) main stream media, it seems to affect very low IQ people, it's like they're transmitting a brain disease almost that affects them. Kind of scary how effective it can be too.
Imagine trusting their outputs on something that can't be verified as easily?
How dhare him calling those LLMs AI
dont trust everything you see on the internet, try it yourself.
ChatGPT
"There is 1 “L” in Google.
G o o g l e → one l."
Gemini:
"There is 1 "L" in the word "google"."
Claude
"One. G-O-O-G-L-E."
since you can have your own instructions for the AI i would not trust any of these "look how stupid AI's are" because you can easily have a setting that makes it lie or act stupid on purpose, if you want that kind of behavior. These channels are quite popular so you can make money of it = not to be trusted.
Chatgpt (openAI 5.5) in typing:
Q How many "L"s are there in the word "google"?
A There are 2 “L”s in Google.
Q How many "l"s are there in the word "google"?
A In the word "google" (g-o-o-g-l-e), there is 1 lowercase "l".
All I did , re-asked the question using lower case l to get the correct answer.
But the first reply was incorrect, and I assure you, @vurt, I didn't instruct it to be dumb.
I am not lying, but then again you don't have to trust me.
Ok. I have all mine set up to check themselves etc. It's the first thing you should do before using any LLM, make it check itself so that it doesn't write something stupid or hallucinate something up, because obviously that is annoying. So it's good to know it seems to be working.
I do think its a bit odd that we have to do this, still. Maybe its because its an extra layer that makes them more demanding , and if everyone actually set them up it would be super taxing. But i don't know.
Also, for spelling or for using it as a timer / make it count etc, while they might be bad at that (at least if not set up) i am not too sure it would be good if people started using LLM's for those tasks. I guess its mostly used by people who wants to find something to complain about, and not because its super useful to have something that can count to 1000, or they need a spell checker (like we don't have those built into windows/browsers).
If people use LLMs to write code, it should at least know how to spell
The numbers don't decide, the system is a lie. The river running dry, the wings of butterflies.
And you may pour us away like soup. Like we're pretty broken flowers.
We'll take back what is ours. One day at a time.
I use it daily for coding and have for the latest one year now apart from a week here and there were i took a break, zero issues with misspelled code. If it did that it would not work at all, maybe a line here and there would be correct but that would not give you runnable code.
Do you see AI's misspell Google or other words people tell it to spell out letter by letter? It doesn't, only for spelling every letter. Do you need AI's to spell out words, no, its useless since its built into windows, into the browsers.
Edit: Asked an LLM to get the correct answer
"A Concrete Analogy
Imagine someone who memorized how to write Chinese characters as whole shapes. They can reproduce 龍 perfectly every time. But ask them to describe each individual brushstroke in sequence and they may fumble — not because they can't draw it, but because they learned it as a gestalt, not as an ordered procedure. The output knowledge and the compositional knowledge are separately held."
What has improved a lot in the recent months for LLM's seems to be memory.. A lot of times these days i'm surprised of the connections it makes from old conversations.
I just asked it for some stats from Metal Slug (how long rapid fire lasts in that game). It directly assumed its something i want to adapt for my game, so it gave me some ideas of how it could work better, it had a good understanding of my game, even though the session was entirely new.
While maybe not exactly crazy that it made the connection (i do speak about my game quite regularly), it's nice that its getting more similar to speaking to someone who knows you a bit and why you would ask something.
Florida sues OpenAl and CEO Sam Altman, the first state to do so,
alleging ChatGPT aided in planning multiple murders and endangers
children. (Ars Technica)
What has improved a lot in the recent months for LLM's seems to be memory.. A lot of times these days i'm surprised of the connections it makes from old conversations.
Which exactly LLM do you refer to? Or, at least, which one do you use most often?
If hes coding I assume Claude.
Not trying to answer for him, just Claude seems to be the best suited, best at it, and best setup to do it with 'cowork', 'artifacts' and 'code' options, outside of just chat.
If not vurt, try Claude. its a good bit better than others at it.
I only use it for small shit I cant be assed to learn like needing custom comfyUI modules and nodes, or quick and dirty npm server apps for small things.
But Ive tried claude, chatGPT, grok, gemini, deepseek, and another I cant remember it sucked.
But claude wins, for me, hands down for code.
ChatGPT is like a high school intern with ADHD and short term memory loss. And its chat gets laggy quick as it dumps all code right into the chat making the chat scroll massive.
Grok is OK, but can sometimes take liberties in "You didnt say not do this, so I'm going to add it, and poorly at that" (I havent tried its new 'build' feature so maybe thats better)
Gemini seems...dunno, milk toast? No good features for it, and compitent but basic at code.
DeepSeek was just bad. It would write code that if I saw it at work I would go "It works, but who wrote this bloated shit?"
Plus Claude (if running local, non browser, version) does: terminal + VS Code/JetBrains, reads/edit files, runs commands, supports MCP, hooks, plugins/skills, subagents, permissions.
Never tried GitHub Copilot. So cannot speak on it.
-We don't control what happens to us in life, but we control how we respond to what happens in life.
-Hard times create strong men, strong men create good times, good times create weak men, and weak men create hard times. -G. Michael Hopf
Disclaimer: Post made by me are of my own creation. A delusional mind relayed in text form.
What has improved a lot in the recent months for LLM's seems to be memory.. A lot of times these days i'm surprised of the connections it makes from old conversations.
Which exactly LLM do you refer to? Or, at least, which one do you use most often?
when i talk to AI's i use chatgpt usually.
when i code i use Codex, VS Code and which model depends on what i am doing.
If i would use the best model for everything then i would be a very poor person right now, it just isn't doable unless you spend thousands. I spend maybe 60 euro a month, and that's ok for a really fun hobby i think, its more fun than any game anyways
"Enlightenment is man's emergence from his self-imposed nonage. Nonage is the inability to use one's own understanding without another's guidance. This nonage is self-imposed if its cause lies not in lack of understanding but in indecision and lack of courage to use one's own mind without another's guidance. Dare to know! (Sapere aude.) "Have the courage to use your own understanding," is therefore the motto of the enlightenment."
i did try some local ones a few days ago, i was not impressed with them for coding, i basically only use Godot now, so i need the bigger models, they do know GDscript.
Signature/Avatar nuking: none (can be changed in your profile)
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
