Don't use ChatGPT. 99% of the requests that free users make go to a useless 8B parameter model. They cannot actually afford to give people real AI.
>>108000783I use claude now because openai is circling the drain and chatgpt gets more incompetent by the day.
>IM ANGRY ABOUT GETTING FREE THINGSit takes a decent gc to run a 8b model. pay up or shut up
Unfortunately, Claude is not available to new users right now. We’re working hard to expand our availability soon.
>>108000783I use Copilot™ now because openai is circling the drain and chatgpt gets more incompetent by the day.
>>10800085110 rupees have been deposited in your account
>complain about gpu shortage>complain that AI companies aren’t providing more computing power for freeIs /g/ retarded?
>>108000783It just works
OpenAI is going to go bankrupt because they can't get funding now that Elon is suing them for unlawful enrichment for switching from non-profit to for-profit.
>>108000783It's way better than any 8b model I've tried. It's better than the 30b models I've tried (but I haven't tried that many, especially recently) .
>>108000783I really don't understand the hype over these AI chatbots in the first place. How many questions does the average person have on a daily basis that can't be answered with a simple Google search that you need to engage a large language model to explain it to you?
>>108000869The shortage is because they own all the fucking graphics cards and they aren't even using them because they don't want to spend more than a penny per 1000 inferences and they think people are too dumb to notice if they just silently swap out their "state of the art" model for fucking llama-2-7b
>>108000906I'm extremely lazy and use it to write ffmpeg codes, though half the time they don't work
>>108000919Please reread the post.
>>108000816>>108000869
>>108000906It's more about it being able to answer questions that take too much effort to answer through Google.For example, you want to convert all webp files in a folder to png. You can just ask ChatGPT to make your a script. Want it to recursively traverse all the subfolders? Just mention it. Want to be able to drag and drop a file onto a .bat file to convert? No problem. Previously if you wanted such a script you would have to know the programs to make it work (ffmpeg or imagemagick) and how to write a .bat file. I need this so rarely that I have to relearn is every time I needed to do it in the past. Now I can just ask AI.
>>108000906writing elmerfem scripts. It still fails 99% of the time if you have anything thats slightly non standard problem. For example deflection of a plate with a body on it every major llm service fails to simulate to generate the script correctly. Just shows how crap elmerfem docs are.
Why would i want chatgpt to have better models then theyll start buying up more shit.
>>108000945This is now the third post that people have made to my question that don't actually address the actual question. Honestly, the fact that people can't read makes so much sense as to why these AI bots are popular. No offense to any of you.
>>108000906googling only gives you unusable seo-optimized noise that was ai generated anyway. the best way to get the information you need directly is through llms, sadly
>>108000906>Why would anyone want to have access to a subject matter expert 24/7Yeah, well, I guess if you are a literal 90IQ moron who literally does nothing and has no ambition to do anything other than sit in his gaymen chair and play vidja games and masturbate all day, then it would seem pretty silly. I bet a lot of things seem silly to you.
>>108000967I didn't mean to upset you
>>108000960>hyper fixating on a specific question >fails to comprehend that normal people read his question as a discussion starter rather than a compulsory question and answer like some sort of examYou’re unemployed, aren’t you?
>>108000960Because your question has an obvious answer. If the number is >0 and the AI can answer that then it is useful.I don't use a vacuum cleaner daily, yet I still have one and find it immensely useful.
>>108000816Is a silver GameCube enough or do I need a purple one
>>108000987I don't know why you're getting so upset. If you think that the average person who browses the technology board of 4chan is "normal", then we're not even on the same planet. The question was very simple and was made in a generally neutral way So as to apply to the wide breadth of people using chat GPT, not just people on this website.
>>108000995>u.ur mad out of nowhere Oof, hit a nerve
>>108000994
>>108001006Ok
>>108000975You think I'm upset because you're stupid? Yeesh. What a stupid thing to assume.
>>108000783>convince it is an 8B model>tell it its an 8B model>it agrees with youWow.
>>108001028Yes. Saying "you are an 8B model" to a retarded response is all it takes to convince an uber-ai that its an 8B "mini" model, and that argument is actually proof of how good this uber-ai is...Really thought that one through, didn't you?
>>108000869Why can't they just optimize their code. They told me AI can do it. They wouldn't just lie to me and investors, right?
>>108001064Why can’t you optimize it and make billions yourself?
>>108001070Because optimizing it wouldn't make me billions. AI can only lose money.
>>108001090Why are you predicting that this is the first time in human history that more computing power is suddenly bad?
>>108000942>say retarded shit>people should just agree because it's about a big companyuh no
>>108001100You're the one saying retarded shit, buddy.
>>108001094Who said it's bad. It would make them lose money slower. But it would still lose.
Wouldn’t the winning move be to use it on a ton of useless shit across many free accounts?
>>108001046>and that argument is actually proof of how good this uber-ai is...I mean, I never said that.LLMs are all garbage.
>>108000869>steal tax payer dollars to make hardware completely unaffordable for the average person>steal source code and art and violate basic copyright law to train the models>reimburse no one while lining your pockets and putting your hand out to take more every step of the way>”hurrrr durrr you think you’re entitled to it fir free???”yes, faggot. i would hope to get at least that much out of this awful bargain i had no say in. anyone with braincells is furious
>>108001291DON'T TALK THAT WAY ABOUT THE BILLION DOLLAR COMPANY. IT'S JUST A SKILLS ISSUE! SIMPLE AS!
>>108001291>steal tax payer dollarsAre you referring to the CHIPS act? Because to me onshoring important high tech industry seems like a decent thing for the government to be investing in, compared to a lot of other shit they spend money on.>to make hardware completely unaffordable for the average personI don't see how this follows the previous claim. Prices have skyrocketed due to demand outstripping supply. Investing in supply chains should help with that, although I imagine it takes years to build.
>>108001028this is hardly a convincing argument in favor of ChatGPT either anon.
>>108000783I don't care how many Bs it has if the output is useful. seek treatment for your autism.
>>108001675>if the output is useful.So, about that...
if free gpt-5 is an 8b model then they're dabbing on open source so hard it's crazy
the gpt chatbot is tuned to confirm user inputs even more than gemini, it may say its 8b model, then it may say its 32b model. if it doesnt provide a clear identification it means it just confirmed your input and you believed that because youre consoooooming retardand now you hallucinate that openai goes bankrupt 2 more weeks. wishful thinking
>>108001835i kinda got suspicious about gemini3 when it replied me its gemini1.5, but that was due to my system instruction insisting on fair identification response and the whole thing that was going through the API, where, supposedly you get the model you select explicitly, not the model that is auto-selected.the GPU retard was previously happily stating multi-modal stuff as good, assuming, probably the perplexity routing and all those retarded open routers that probe user input complexity and select model for the developer. no serious development may be implemented on those, only dumb re-selling of their chatbots.so always be suspicious about initial text, if you didnt ask for explicit clarification, it will usually contain "good" or pleasing parts for engagement. stucturally they got engagement first and cta last, in the cta it kinda concludes and asks you "do you want to know more?"
>>108000783Bro is getting mad at a chatbot
>>108001432Things never become cheap
>>10800078360 more bils from Nvidia, Microsoft, Amazon, 30 bil from SoftBank, 50 bil from Saudi
>>108000783NoUse ChatGPT as much as possible, they lose money everytime you do that
>>108000783Cope, brown. aiGODS are replacing you
