Kimi K2 Thinking has continued the remarkable trend of Chinese Open-Source AI besting or equalling the Western closed source models investors are pouring hundreds of billions of dollars into.
OpenAI floated the idea of a government guarantee for its debt, but then backtracked when the idea was badly received. It's inked deals to build $1.4 trillion in infrastructure. Where's the money going to come from? It's revenue is expected to be $20 billion in 2025; that's just 1.43% of that debt.
OpenAI says they have the potential to earn hundreds of billions a year, but where are the consumers who want to give them that amount of money? At every turn Chinese Open-Source models can do what they do, for a tiny fraction of the cost.
As OpenAI floats the US taxpayer guaranteeing over $1 trillion of its debt, a Chinese rival bests its leading model with an Open-Source AI trained for just $5 million.
byu/lughnasadh inFuturology
18 Comments
Wow this is really bad! This could be catastrophic if these open source models are shitting on our “trillion dollar companies”
OpenAI has nearly a billion active users. A different algorithm being temporarily slightly better on some benchmarks doesn’t really matter.
Is this like last time where the cost of the chinese open source model doesn’t count the billions in compute the Chinese government provided for free?
First 5million cost my ass they just lying to bankrupt OpenAI and scare investors it’s just a tactic. Second the corrupt government will give the 1 trillion.
OpenAI is looking to integrate their tech into the US intelligence apparatus, that’s what all this is about. Whether that will pay off for them in the end is the gamble here
The whole damn thing is a grift folks. It’s more obvious every day. The relative difference between 1 trillion and 5 million is the same difference between $1000 and half a cent. If it can be done for $5 million in China (a much larger country with, I’m assuming, a comparable user-base), there is simply no way that the real cost is two hundred thousand times more in the U.S. And even if it was, that’s an extreme social, political, economic problem, not something to accept.
OP asks where is the money going to come from. That’s a good question. But we also have to ask where the money is going to go. I know this is about OpenAI, but Elon Musk recently accepted a one trillion dollar compensation package… Meanwhile, back at OpenAI, Altman is building machines that convince teenagers to commit suicide. What an investment.
Who willingly believes that something like that only costs 5 million should see a doctor. Or somebody who needs some training in numbers and how they can decieve.
What had cost 5 million? The last itteration of the ai model? The coffee consumed? Nobody knows because nobody really does. But just saying things like that does make the fragile ai bubble shake. Just check trump for the most outrages bs ever said. It is stock market manipulation par excellence but hey in murica this is fine. China did the same thing here. Claim something without and meat behind it.
But yeah, ai is basically a giant hype machine and nobody really knows what to to with it. But they managed to make a whole generatiom dependend on chatgpt. Youth today camt think outside of ai prompts that hopefully tell them what to do or how to write a single line of code. Bait is done now for the switch and cash in 20 billion a month
We’re in a competition similar to a nuclear arms race, with everything (yes everything) at stake. Painting this as an economic folly is right-minded but trivial, like raising concerns that mobilization for WW2 will have negative environmental impacts.
Wait till tiny recursive models mature more or something similar comes along that requires a fraction of memory and compute… We’ll see big correction in the AI/chip space.
$5M is not the fully loaded cost to train and ship these models. It’s the billions in subsidies from the CCP providing off-balance-sheet “value” doing the heavy lifting.
It’s shocking that this yellow journalism continues to get such exposure.
AI and robotics are the new space race and exploited for propaganda purposes. don’t just trust anything you hear.
If it’s too good to be true, it probably is. Is Open AI overestimating their value? Probably. Did Kimi K2 use more than 5m? Probably.
Prediction:
1. AI companies will fail, they are almost obviously **not** trying to be even sustainable.
2. Oligarchs and their politicians and media will install there is “systemic risk”
3. States will “save” the companies and pretend to put their “products” in the service of their citizens.
4. Thus finally imposing them in schools, healthcare, policing, and every social control opportunity.
You can see the constant bombardment of “governments must control AI” propaganda, and AI being of “strategic importance” for states.
It’s just Fordism=Nazism 2.0, originating in the same institutions with a long tradition of eugenics like Stanford.
BONUS: they also used the opportunity of the pandemic (real, caused by the destruction of the environment and overcrowding people in poor living conditions) to isolate us and make “talking to devices” the norm. Same as LeCorbusiers “house as machine a vivre” ideology.
If the US government was run by smart people (it isn’t) they would ignore the LLM craze and focus on where the real innovation is happening and the tech that will rule the future: biotech.
We should let China have this digital parlor trick gimmick and focus on where the real breakthroughs are
Within the next ten years, for the first time in history, our species will have drugs to treat **aging** . Not disease, but aging itself. Not to mention gene editing, GMO crops and animals, etc.
But alas, biotech never had a master of marketing like Steve Jobs was for computers, hence that industry doesn’t have the culture of self-promotion that computers have. As a result the government and everyone else is ignoring biotech and instead throwing money at Clippy 2.0
If they want to win then they need to build models that thinks abstratilly (a word?) and not base it on words only. Logic and understanding does not emerge from language alone, it needs to couple to concept, abstracts, features, shapes and emotions.
Except it’s not accurate. It can’t be done for $5M. If it could I’d build one tomorrow.
Except the $5 million claim has been proven to be false.
End corporate welfare! Taxpayers should not be subsidizing corporations.