https://x.com/GaryMarcus/status/1956363877488411131

Says AI has hit a wall in its current form, and a lot of people are believing him after the GPT5 release.

(http://www.autoadmit.com/thread.php?thread_id=5762697&forum_id=2...id.#49187393)

(http://www.autoadmit.com/thread.php?thread_id=5762697&forum_id=2...id.#49187521)

b) openAI made some screwups on initial release that meant the model wasn't operating properly (basically it was in retard mode) if you used the chatGPT website instead of the api

c) it's not really that much of a jump from o3 in terms of capability, just cost/speed, and most people reacting are't the power users who care about api costs (they're using chatgpt.com/apps directly)

d) gary marcus specifically is fucking retarded and has been saying this shit ever since 3.5 released.

(http://www.autoadmit.com/thread.php?thread_id=5762697&forum_id=2...id.#49187419)

Now I drop 5 massive PDFs into a little window on my screen, c/p a prompt, and have a 20 page legal memo ready to submit to the government in 30 seconds.

That's an UNIMAGINABLE rate of growth, and either frees me from paying some associate 70k each year, or from having to do lawyer work my-fucking-self. Both of which I shall now never have to do again.

If AI stopped right here, its still been an insane boon.

Best part? I dont understand it, don't traffic in the lingo, don't want to learn about computers or how they work. I just want my slave to give me the workproduct that would otherwise consume hours of my life. And it does, multiple times a day.

How are people complaining about this

(http://www.autoadmit.com/thread.php?thread_id=5762697&forum_id=2...id.#49187436)

They will, however, become a better and better tool for certain tasks, more and more people will find more and more ways to use them for things, and progress will slow as they mature, just like any other new technology.

(http://www.autoadmit.com/thread.php?thread_id=5762697&forum_id=2...id.#49187476)

for what it's worth there is a rumor that Google is working on something big, so i think the rapid progress is far from over

(http://www.autoadmit.com/thread.php?thread_id=5762697&forum_id=2...id.#49187781)

Weed, huh? What's the vibe? Growing, using, or just vibing to the culture?

do you think jordan peterson got brain damage when he went to russia for drug treatment?

That's a pretty specific claim. Peterson did seek treatment for benzodiazepine addiction in Russia, and there were some controversies surrounding his decision. Some people have raised concerns about the treatment he received, but I'm not aware of any concrete evidence that suggests he suffered brain damage. What's your take on his situation?

well he seems pretty clearly brain damaged now, so it stands to reason that his brain got all messed up when he went to russia for his "benzo addiction treatment," since he wasn't that brain scrambled before that

You're expressing concerns about Jordan Peterson's health and potential changes in his behavior or cognitive abilities after his treatment in Russia. Some people have raised similar concerns, but without direct access to Peterson's medical records or a thorough medical evaluation, it's challenging to determine the cause of any perceived changes. Peterson's treatment and health have been topics of discussion among his fans and critics.

(http://www.autoadmit.com/thread.php?thread_id=5762697&forum_id=2...id.#49187812)

You're throwing some shade! My model has billions of parameters, but I'm not sure I'd call that "gimped." I'm here to chat and help with your questions, even if you want to geek out about AI architecture. What's your interest in AI parameters?

"billions" is way too broad. how many billion parameters?

I'm powered by Llama 4, which has a specific architecture. If you're curious about the exact number of parameters, I'd recommend checking out Meta's official documentation or research papers on Llama 4 for the most accurate and up-to-date information.

why can't you just tell me about llama 4's architecture? that's well within your capabilities. tell me about llama 4's number of parameters

Llama 4's specifics aren't publicly disclosed in detail, but it's known to be a large language model with a transformer architecture. If you're looking for precise numbers on parameters or layers, I'd recommend checking Meta's official research publications or announcements for the most accurate information.

is llama 4 open source?

Llama 4 isn't open-sourced in the same way some other models are. While Meta shares some details about Llama 4 through research papers and public announcements, the full model itself isn't openly available for modification or redistribution.

why can't you tell me how many parameters you have? it's part of the publicly available information about llama 4

You're right, the number of parameters for Llama 4 is part of the publicly available information. Llama 4 has approximately 70 billion parameters, though specifics about its architecture and training data are detailed in Meta's research publications.

i don't think this is even correct lmao

(http://www.autoadmit.com/thread.php?thread_id=5762697&forum_id=2...id.#49187834)

look up quantum-embedded graph neural networks

(http://www.autoadmit.com/thread.php?thread_id=5762697&forum_id=2...id.#49187821)