- cross-posted to:
- usa@midwest.social
- cross-posted to:
- usa@midwest.social
You must log in or register to comment.
Good.
No wonder they’re panicking so hard over open weight Chinese models
i wish them ruin, you made my day
Repeating the DotCom burst.
Idk man, to me, it feels like we are further from AGI than fusion reactors actually being hooked up to the grid, considering LLMs are still just glorified autocorrects, needing a ton of prompt injections from their providers. And yet, the AI companies make it seem like it’s right around the corner (obviously, because $$$), whereas to me it feels like we are still at least 50 years away. Maybe AGI will be the next fusion generator chase.
Ironically there was notable news in the fusion reactor space this week as Gemini 4, building on the work of Matt Landreman and GPT-6 Astra a little while ago, expanded counterexamples to Grad’s conjecture which had claimed exact solutions for controlling a key aspect of fusion were impossible.
So fusion may actually be more around the corner thanks to the advances of capabilities of AI which helped prove that the fusion problems aren’t as hard as they were thought to be for decades.
AGI has been achieved more or less because it just means self sufficient at completing tasks. An LLM is just a component these days because they are Vision-Audio-Language-Models with multi frequency ‘thought’ loops. The next letter is A for action where they just have high frequency outputs for direct computer and robot use.
Just because it is bad AGI doesn’t mean it is not AGI right
$8B real loss on $16B revenue. ($72B is a bondholder gain (accounting loss) from conversion). $9B in training costs not counted as loss, even though it is worthless whenever next model version is released.
$561B in forced spending on compute over next 10 years, with $750B total compute costs lined up. Structural losses are baked in for a few years.
But, when the word guessing program gains consciousness, it will be capable of doing the job of every worker in the world, and eager to do those jobs, and anybody who owns Anthropic shares will be a trillionaire in an economy that totally works and is healthy. So, better invest now!
Musk giving 1 share (about $160) to all US children with a Trump account, will make the next generation cheer for skynet genocide and oppression. “For a while, it added 10-20% to my net worth.”
Stop_trying_to_make_AI_happen.jpg
Why haven’t they just asked Claude how to make money with their business model? /s
They have.
-
Wait for excited CEOs to fire all their staff.
-
Crank up the prices.
2 only works if regulatory capture in place to prevent cheaper competition from open models.
-
“A MINOR SETBACK ON THE PATH TO THE SINGULARITY!” -
<ADAM DRIVER> MORE! </ADAM DRIVER>
[…]included a $72 billion accounting charge reflecting an increase in the estimated value of bonds that could eventually convert into Anthropic shares, rather than cash costs tied to its operations, the document said.
So dept isn’t cash cost that’s tied to operations? All the debt did was finance the operations. Reading this paragraph and you’d wonder: who in the hell issued those bonds in the first place?
Repaying debt is cash cost. But this is convertible debt that could be converted to stock, in which case the holders of those newly converted shares don’t get paid cash from the company, but instead own a chunk of the company worth more than that cash. So the conversion itself would alleviate the cash demands of the company, while also diluting the ownership stakes of any preexisting owner. That’s what that paragraph means.
Companies issue convertible bonds usually because they can get better interest rates on it, in the event that they need to repay it. Or they’re happy to deal with diluting their shares if the company becomes so valuable, because that means they’ve gotten a lot richer and just have to share their newfound riches with the lender.
included a $72 billion accounting charge reflecting an increase in the estimated value of bonds that could eventually convert into Anthropic shares, rather than cash costs tied to its operations
It’s a very common way for startups to raise money. In this case, the bond value goes up because of a perceived increase in the value of Anthropic.
The ai crash of 2026 can’t come soon and fast enough
The RAM cartel is figuring out a new excuse for a price hike before allowing it.
Every local food truck operator makes more than they do, and for good reason.
Got a local food truck epidemic of mid overpriced food and no food safety standards, in my college town it was great, back home makes me wish they regulated more
Cuz they got churros.
It is obvious to me that AI computing is a COMMODITY, in that it will inevitably be something that can not be effectively gate-kept.
These two major US companies (OpenAI and Anthropic) are playing the game of “locking in” a monopoly in a product category that has yet to mature. This is a pretty old capitalist game, of course… Uber, Microsoft, facebook,and many others have played this game in the past.
But, unlike OSes or word processing file compatibility I don’t see a “network effect” in that an open source/weight model can/will do 99% of what a proprietary model can, will do it “transparently,” will eventually become so optimized that it can run locally, etc. etc.
I mean: maybe its name recondition that is worth “trillions” but a race to be “Xerox” or “Kleenex” or even “PC” (as in IBM PC) did not really work out well for these companies.
I think that even the established players (MS, Google, etc.) are biding their time, making sure that they don’t get locked out while non-proprietary solutions become feasible.
I means I know that much of the issues have to do with the limited availability of hardware, but locking down supply is not sustainable for the “big 2” and sooner or later more adequate memory and GPU supply will enter the market…
I dont think a local model will do 99% of what the big ones can, and even if it could, it would require an enormous amount of memory and cpu to do it.
People pay for a good user interface and convenience, that’s well known. Netflix and Spotify are built on that idea and people starting using them even over free pirated stuff :)
Maybe not “today” but today is still early days.
I’ve been consuming lots of “build your own local AI coding agent” kind of content lately. No doubt this is the same as the guys that are welding together their own versions of Shelby Cobras and stuff… but in a few years you’ll be able to buy a “server” with enough GPU and ram to host strong open models and serve them to your little office.
I work in a small company, and if AI plays out the way Anthropic sees it’d their AI bill will be $100k every year (fire 2-3 people, pay for AI so that the remaining people do more). At that level of expenditure, why wouldn’t my poss pay as much as a copy machine $25k, lets say, go give “local AI” a chance?
The stock valuations that most tech companies carry (and certainly the 2-3 trillion that is being talked about for Anthropic) speak to potential future revenue.
To me, I see computing and actual “model” advances meeting the need within a few years… if not by entrenched companies like Nvidia, then by insurgent chipmakers or new (Chinese or otherwise) chipmakers.
In 10 years you’ll have “opus” is not better running in your phone and probably something more powerful in a PC.
Qwen3.8-flash-next broke the vram requirement with its new architecture (which is a preview of their 4.0 architecture). You can run a 177b parameter model on 12gb VRAM, 64gb ram and 80gb SSD at 30tok/s and the speed scales almost linearly with increases to vram.
Incidentally, this was proven at almost the exact time all the calls for “regulating local models because they’ll spread into the toasters” started.
You absolutely can have local models that run on consumer grade hardware, and the proprietary companies know it and are scared to death.
Interesting.
Are these really able to be as good as chat gpt? Because every time I’ve tried a model like that, it has not been able to be nearly as good. I have 48 gb memory on my machine so I can’t run this particular one though.
I would buy more ram and a whole new computer if the market wasn’t insane.
i would not recommend using the SSD to run local LLMS,its slow
best: VRAM only
good: VRAM + RAM
bad: SwappingI don’t think they’re insinuating running inferences from the SSD. But you need somewhere to store the weights for the GPU to load.
i am trying to say: you need the SSD to store it but not swap to it due to its slowness
I won’t pretend to be an expert but my understanding is that the architecture loads experts from the SSD on demand with the RAM serving as a hot cache and the VRAM running the active experts.
As I said above I was able to run 70tok/s on my 4090w64GB ram and a samsung pcie4.0 ssd.
I haven’t run the benchmarks myself, but I can say the quality of response I was getting seemed on par with the frontier models I use for work (Claude opus/fable) and the published benchmarks for the model compete with those same models.
It is a q3 quantization so there is some small loss of accuracy, but from what I’ve understood those are modest losses in the worst case.
There are models that don’t need to consume the entire worlds data infinitively. We’ve had then before llms, things like the pi ai camera can do ai processing on image with something like 8mb of on board storage.
The problem is anthropic, open ai, who ever are trying to do the stupidist approach and assume they can brute force money and power at the problem and don’t understand that they need exponentially infinite data in a way that is unobtainable.
In an ideal world there would be a million little small local models that so specific tasks and can be ran on consumer hardware.
maybe they can do like 60% of the big ones?
I dont know, maybe, but I think either way, local models will be a fringe interest for technical people, and not something that will affect the income for these big tech Ai corps in any significant way.
Because you simply can’t run a model with all the worlds knowledge on your home computer.
Someone said that maybe there will be small specialized models that only know about a certain thing, and sure, that’s possible.
You also can’t run a model with all the worlds knowledge on a supercomputer. That simply doesn’t exist. To think that even the largest frontier models can do that is a mistake.
The real question is: will the median user be able to run basically everything they need locally, and how much compute would they need to do that? These companies are betting the farm that it will never happen, and that they’ll be able to charge whatever they want in rent to a captive audience.
But I think these tech CEOs and ML researchers talking about “AI killing us all” is related via the aphorism: It’s easier to imagine the end of the world than the end of capitalism. They can see the writing on the wall, but can’t imagine a future for themselves.
All those loans are getting hot. They’re starting to toss them around and trying to hide them.
I have strong doubts they can ever make this particular approach to AI work at all, let alone in time to cover their bets.
This path looks like a dead end. But, I don’t know jack about shit.
The issue is they are too expensive, it’s not really a limitation of the technology, well it might be, but that’s not the issue here. They are paying Nvidia Billions for their entire 2026/27 GPU runs and they can’t get the customers to cover that. It’s like $200 pm for their max plan. It’s almost cheeper to hire a human in India.
There’s a reason so many AI projects have turned out to be Actually Indians.
🐤
My monthly salary as a software developer is 80 USD
I hope it pops
