Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
OpenAI Alleges DeepSeek Used Its Models for AI Training
- Thread starter MacRumors
- Start date
- Sort by reaction score
You question implies there was a second wrong
I'm betting Microsoft are quietly biding their time for OpenAI's stock to tank so they can buy up the company, throw out the high running costs and use the pieces to build an on-device LLM directly into Windows 12.
Half the reason investors are spooked is because Altman et al have largely been taking them for a ride. Consumers won't pay for AI when they're used to getting a version of it 'for free', think Siri or Alexa. OpenAI need VC cash to keep the lights on, cash they are literally burning through that nobody will ever see a return on. If they see a Chinese company can build a good-enough version for $6m why would they ever continue to invest in ChatGPT?
Nvidia have been overvalued for years based on an easily-popped AI bubble and the eternal scam of Crypto Currency
Half the reason investors are spooked is because Altman et al have largely been taking them for a ride. Consumers won't pay for AI when they're used to getting a version of it 'for free', think Siri or Alexa. OpenAI need VC cash to keep the lights on, cash they are literally burning through that nobody will ever see a return on. If they see a Chinese company can build a good-enough version for $6m why would they ever continue to invest in ChatGPT?
Nvidia have been overvalued for years based on an easily-popped AI bubble and the eternal scam of Crypto Currency
They are actually much worse
They are the ones that did the pillaging in the first place
Yes, every LLM carries the bias of the data it is trained on. I can trust ChatGPT to answer any question with a western political bias. However, there is a difference between bias and censorship.
This difference is important to me because I use LLMs as a substitute for search engines. Here's another interesting question one might ask "Characterize China's current monetary policy". ChatGPT 4o gives me a four point answer with references to back up its answers and links to other articles about China's commitment to spur economic growth in 2025. DeepSeek replies with: "I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses."
Last edited:
Sure. My point is, everyone knows how a Chinese model will answer (or skirt) questions about certain historic events. The more subtle biases are much more interesting. And western models will be full of those as well.
Just wanted to second (or Nth) all the comments about Open AI, having stolen IP from virtually everywhere they could get it, now complains that someone used query responses from their own system to build their own model.
Ha ha f**king ha. All crooks.
Ha ha f**king ha. All crooks.
I know the difference between bias and censorship. I just don't think it's surprising or interesting. In fact, it's the most predictable thing about a Chinese AI service I can think of.
Last edited:
I read a neat article yesterday about how the Deepseek programmers used PTX instead of CUDA for their coding because CUDA (a higher level language than PTX) didn’t expose certain functions/methods the Deepseek developers used to improve performance.
DeepSeek's AI breakthrough bypasses industry-standard CUDA for some functions, uses Nvidia's assembly-like PTX programming instead
Dramatic optimizations do not come easy.
I agree. They’re good aren’t they.
What was the line in Pirates of Silicon Valley? Bill said he got to the TV first, Steve?
It’s also incredibly hard to be sympathetic when, in addition to the first point about stealing what is stolen, they released this as open source, allowing anyone to use it or develop with it, which ironically is what OpenAI was supposed to do from the outset.
I’m sorry, but I’m just not feeling bad for the greedy corporations in this case. If equities take a hit, so be it. I hope this ushers in many more open, non-proprietary AI models and democratizes the tech. The “tech bros” have consolidated MORE than enough money and power. If this undermines that a little, I’m all for it. I don’t want to live in a dystopian future where we are all under the control of Skynet.
I want to live in in a dystopian future where we are all under the control of SkynetS!
I’m sorry, but I’m just not feeling bad for the greedy corporations in this case. If equities take a hit, so be it. I hope this ushers in many more open, non-proprietary AI models and democratizes the tech. The “tech bros” have consolidated MORE than enough money and power. If this undermines that a little, I’m all for it. I don’t want to live in a dystopian future where we are all under the control of Skynet.
I want to live in in a dystopian future where we are all under the control of SkynetS!
They’re all crooks and the norm seems to be; say one thing, do another, deny or apologize when caught and complain when you’re the victim of your own deceitful behavior. 🤔🤔
Not coding, but using company-proprietary data that needs to be kept private. The solution may be convincing my boss to sign up for an enterprise account of ChatGPT, but think he'd be more comfortable with it living locally (but not comfortable enough to buy me a new Mac 😂) .
Oh the irony, so much irony.
Not defending DeepSeek or any other one of the recent AI startups, they're all at fault for the approach with which they trained their models to become relevant and get funding to do more of it.
DeepSeek is only exposing the AI boom for what it is: a lot of unethical (and very expensive) smoke and mirrors.
Not defending DeepSeek or any other one of the recent AI startups, they're all at fault for the approach with which they trained their models to become relevant and get funding to do more of it.
DeepSeek is only exposing the AI boom for what it is: a lot of unethical (and very expensive) smoke and mirrors.
They're surely not complaining about using other peoples copyrighted material or intellectual property are they? Cause that would break irony meters around the world.
OpenAI Believes DeepSeek ‘Distilled’ Its Data For Training—Here's What To Know About The Technique
White House AI czar David Sacks alleged Tuesday that DeepSeek had used OpenAI’s data outputs to train its latest models through a process called distillation.
www.forbes.com
Starting to see a few online news running with this insanity news topic.
Mike Masnick, the founder of the tech news outlet TechDirt, commented on OpenAI’s reaction to distillation: “So, look. I'm sure I'm in the minority here on Bluesky in believing that training AI systems isn't copyright infringement. But, also. Dude. There's no way OpenAI can make this argument without looking very, very silly.”
Last edited:
