Oh sweet baby Jesus no, burn AI to the ground, humanity can't be trusted with it

[Chimpzy]

Half a trillion dollars on an AI boondoggle. That's "fiscal conservatism"!

It's not being paid for with taxpayer money. Funding is private from banks and such. That said, you might end up right after all when this scheme fails and the culprits are suitably rewarded with government bailouts

Speaking of fail

https://www.axios.com/2025/01/22/stargate-elon-musk-trump-altman-openai-project

Something is salty

 

[Agema]

It's not being paid for with taxpayer money. Funding is private from banks and such. That said, you might end up right after all when this scheme fails and the culprits are suitably rewarded with government bailouts

Given it's a private initiative, this could be a business plan that has been in the works a long time and they've just rolled it out via Trump as a way to curry favour.

If it really has been stitched up last minute with some political meddling, chances are higher it'll be one of those plans that they can readily back out of, go slow, and generally leave to die after an outlay that's a miniscule fraction of the initial promise. If that's the case, no-one would ever really notice: no big announcement, no headlines, just a vague memory that once some company said that they would do something.

 

[hanselthecaretaker2]

Trump aides ‘furious’ with Elon Musk for trashing $500B AI project: ‘He gives zero f–ks’

Gotta say, Altman and his corporate alliances scare me quite a bit more than Musk, all things considered.

 

[The Rogue Wolf]

Leonard French delves into Microsoft's condescending reply to a legitimate worry about CoPilot's effect on client confidentiality. Note that Microsoft will not let users disable CoPilot, let alone remove it.

 

[XsjadoBlayde]

Spoiler: Son of a

 

[Dirty Hipsters]

The world is freaking out about Deepseek, a new Chinese AI with backing from the Chinese government. Nvidia stock price dropped like a rock with the biggest market loss in history losing nearly $600 billion in value.

And I'm sitting here going "it's about fucking time." Nvidia is massively overvalued. AI is massively overvalued.

CEOs are throwing money at it and drooling at the prospect of replacing all of their employees with AI. No more lunches or bathroom breaks, or work-life balance, or sexual harassment lawsuits, the AI will do it all! meanwhile AI can't do basic math, can't tell you how many Rs are in the word strawberry, and constantly makes up fake information. And the more AI content there is on the internet the worse AI becomes because it gets trained on AI content becoming more and more subpar. The whole thing only works when there's human work for the AI to copy, steal, and exploit. Brilliant.

 

[Agema]

The world is freaking out about Deepseek, a new Chinese AI with backing from the Chinese government. Nvidia stock price dropped like a rock with the biggest market loss in history losing nearly $600 billion in value.

And I'm sitting here going "it's about fucking time." Nvidia is massively overvalued. AI is massively overvalued.

Yep.

I am fascinated that this Chinese outfit appears to have similar capability at much lower cost and processing requirements. I cannot help but wonder if the US tech titans are flabby: given vast investment resources, they haven't needed to be efficient, because they can just hose money at everything. Chinese developers, with lower budgets and facing chip bans / restrictions and other limitations, have had to be smart rather than employ brute force.

It's a real smack in the face for the US tech industry. If they're more expensive and can't do better, then it's just a matter of time until they're history.

CEOs are throwing money at it and drooling at the prospect of replacing all of their employees with AI. No more lunches or bathroom breaks, or work-life balance, or sexual harassment lawsuits, the AI will do it all! meanwhile AI can't do basic math, can't tell you how many Rs are in the word strawberry, and constantly makes up fake information

And this is the painful thing. AI companies to a certain extent have literally no interest in whether their product does the job better: they're going to lobby to have businesses replace their staff with AI just so they can make money. They certainly have no interest in what the cost is for the people their product replaces.

And the more AI content there is on the internet the worse AI becomes because it gets trained on AI content becoming more and more subpar. The whole thing only works when there's human work for the AI to copy, steal, and exploit. Brilliant.

Yes, this is the fascinating thing. Studies suggest that LLM AIs may end up destroying themselves; as they increasingly create content, later LLMs will increasingly be trained on content generated by earlier LLMs, and each generation of output becomes more degenerate than the the last. Having a "pre-LLM" database will become crucial, and finding a way to weight that in training to prevent degeneracy.

 

[Chimpzy]

https://www.ft.com/content/a0dfedd1-5255-4fa9-8ccc-1fe01de87ea6

OpenAI says it has found evidence that Chinese artificial intelligence start-up DeepSeek used the US company's proprietary models to train its own open-source competitor.
The San-Francisco-based ChatGPT maker told the Financial Times it had seen some evidence of "distillation", a technique used by developers to obtain better performance on smaller models by using outputs from larger, more capable ones. This allows them to achieve similar results on specific tasks at a much lower cost.

"We know [China]-based companies — and others — are constantly trying to distil the models of leading US AI companies," OpenAI added in a statement. "We engage in countermeasures to protect our IP, including a careful process for which frontier capabilities to include in released models, and believe as we go forward that it is critically important that we are working closely with the US government to best protect the most capable models from efforts by adversaries and competitors to take US technology."

One person close to OpenAI said that distillation was a common practice in the industry and highlighted that the company offers developers a way to do this using its own platform, but said: "The issue is when you are doing it to create your own model for your own purposes."

OpenAI is currently battling allegations of its own copyright infringement from newspapers and content creators, including lawsuits from The New York Times and prominent authors, who accuse the company of training their models on their articles and books without permission.

Lol. Lmao even. AI tech bros complaining about being stolen from is never not funny.

 

[Casual Shinji]

The world is freaking out about Deepseek, a new Chinese AI with backing from the Chinese government. Nvidia stock price dropped like a rock with the biggest market loss in history losing nearly $600 billion in value.

And I'm sitting here going "it's about fucking time." Nvidia is massively overvalued. AI is massively overvalued.

CEOs are throwing money at it and drooling at the prospect of replacing all of their employees with AI. No more lunches or bathroom breaks, or work-life balance, or sexual harassment lawsuits, the AI will do it all! meanwhile AI can't do basic math, can't tell you how many Rs are in the word strawberry, and constantly makes up fake information. And the more AI content there is on the internet the worse AI becomes because it gets trained on AI content becoming more and more subpar. The whole thing only works when there's human work for the AI to copy, steal, and exploit. Brilliant.

Yeah, but they'll make money off of it so long as the scam persists. And once the bubble bursts they find a new one to inflate.

And the billionaires lived happily ever after.

The End.

 

[Agema]

Lol. Lmao even. AI tech bros complaining about being stolen from is never not funny.

Yep. It's right up there with pocketing zillions in public money (grants, subsidies, tax rebates, piggy-backing off publicly funded research) and then demanding we bow down before their entrepreneurial genius.

Of course, I think Altman's point isn't actually about copying, it's more fundamental, about asserting the superiority of OpenAI's technology. What he's saying is that DeepSeek could only get there by hitching a tow on OpenAI, thus its accomplishment is superficial rather than substantive - and therefore investors are still best off backing him.

OpenAI is currently battling allegations of its own copyright infringement from newspapers and content creators, including lawsuits from The New York Times and prominent authors, who accuse the company of training their models on their articles and books without permission.

So, interesting one. Authors were quite pleased that the publishing industry fought back against AI for using their books as training. Except it turns out that publishing companies have been pushing new contracts regarding AI. We have to remember that business is inherently about screwing everyon else for money, and that's how publishers feel about authors, too. It seems that publishing firms don't actually want to defend the ability of human authors to make a living from being ruined by AI-generated content, it's that they want to be the ones who control and profit from the AI-generated content that ruins the ability of human authors to make a living.

 

[Chimpzy]

Of course, I think Altman's point isn't actually about copying, it's more fundamental, about asserting the superiority of OpenAI's technology. What he's saying is that DeepSeek could only get there by hitching a tow on OpenAI, thus its accomplishment is superficial rather than substantive - and therefore investors are still best off backing him.

Are they tho? If I was an investor I would be interested in knowing, if DeepSeek managed to create an ostensibly better, more efficient product using less resources in the way they did, then why didn't OpenAI do the same, them being so cutting edge and innovative and all?

But that may be giving investors too much credit.

 

[Agema]

Are they tho? If I was an investor I would be interested in knowing, if DeepSeek managed to create an ostensibly better, more efficient product using less resources in the way they did, then why didn't OpenAI do the same, them being so cutting edge and innovative and all?

But that may be giving investors too much credit.

Imagine that a new web search engine ("Blob") worked by putting searches through Google, DuckDuckGo, etc. and then applying an algorithm to refine their collective outputs to provide to the end user. It might be a better result for the user, but actually most of the work is being done by the other search engines. Without access to the other search engines, "Blob" would be pretty much useless. The implication therefore is that DeepSeek is only any use because it's used OpenAI to do lots of the work, and thus without OpenAI, it's a bust.

OpenAI however can't do the same itself, because this model works by piggy-backing off someone else's hard work... as OpenAI is the one doing the hard work, it's got no-one to piggy-back.

Presumably, OpenAI could make a parallel "light" AI and train it off their main AI to achieve the same as DeepSeek - with lower resource cost. However, presumably that would also retard the development of their main AI, because then it's no longer being used to collect inputs and generate outputs to drive development.

* * *

One might argue it's not so bad if AI ends up as there being sort of core, "heavy" AIs doing the main development, and then a load of "light" AI interfaces that use these as a springboard to create faster, better, less resource-intensive consumer output. But you'd certainly expect those "light" AIs to make agreements and pay the "heavy" AIs for providing so much of the background work. Just like many think any AI developers should pay human creators for the content used in their programming.