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
Not counterfeit....Open source with a Chinese development. Can we stop this red scare xenophobic rhetoric every time we get a story of someone from China doing anything.
Oh please. Among other issues, ChatGPT was trained on data that is behind paywalls. Hopefully I don’t have to explain to you why that is a problem.
Edit: And by the way, I know you asked someone else and not me, but no I’m not “only now” concerned with people copying my website. First time I called my lawyer due to someone copying articles on my website was about twenty years ago.
Last edited:
Just to be clear to everyone: No one is running DeepSeek locally on a MacBook. What you can run on a MacBook are distillates of DeepSeek that are substantially less capable. I do this myself, they are useful for certain tasks.
Good for you. ChatGPT was trained on data behind paywalls, which is equivalent to scraping behind your “doors”. Also, just because something is available online does not mean the owner allows you to do whatever they want with it. Would you be perfectly happy with competitor websites totally copying your online sales pitch to steal your customers?
Last edited by a moderator:
Sorry, I should have been more clear. Yes, I'm running much simpler version of DeepSeek. That said, I just found it interesting that my local copy answered a question about Tiananmen Square while the online version does not.
You’re stealing our stolen content! No fair!
I'm assuming the online versions are doing some post-processing that the local distillates are not. I doubt they try to train the models to do this up front.
This reminds me of Apple screaming at Microsoft that they stole their GUI, while Xerox looked on at both of them.
What open AI alleged has nothing to do with IP. At best it’s terms and conditions. Open AI is in serious cope mode.
Who cares? OpenAI used copyrighted books, movies, photos, everything on the internet to train their model.
Why do they think they should get copyright protection but no one else does? Truly a perfect time to break out the microscopic violin
Why do they think they should get copyright protection but no one else does? Truly a perfect time to break out the microscopic violin
Yep. It’s not apples to apples. But it’s like opening up App Store for smart phones in 2009. I am already seeing zero shot models trained for under 40 bucks using the deep seek.
Apple did not steal from Xerox. They made a deal which turned out to be very one-sided because Xerox didn’t realize what they had. Very different from stealing.
UNLIKE in America, where our AI companies are required to respect all copyright and IP laws and pay for their training data. Oh wait…
A lot of people in this thread seeing "OpenAI Alleges DeepSeek used Its Models for AI Training" and assuming that means "DeepSeek stole OpenAI models, changed one or two things, and called them their own" when in fact it means "DeepSeek used OpenAI as a tool to make its models better".
"websites totally copying your online sales pitch to steal your customers"
Has it been publicly available for anyone? If so, who cares?
OpenAI and the other LLM thieves should be regulated out of existence, with their decision makers subject to criminal prosecution. They’ve built their business on stealing other people’s copyright works.
OpenAI have no moral argument against anyone who steals from them what was built on theft.
OpenAI have no moral argument against anyone who steals from them what was built on theft.
Using the output from a "trained model" to train another model is the issue here. That's technically theft. The investment went into training the original model to begin with. Training is a highly-involved and expensive process, and it sounds like DeepSeek might have jumped the line. I'm not surprised if that's the case, but I'm really hoping it's not true.
Different situations. OpenAI did the training from "suspected" immoral sources. That training was the hard part.
It sounds like DeepSeek took OpenAI's hard work and bypassed the training process entirely. Totally different from siphoning data for the original training, of which OpenAI may be guilty of indeed.
Two different suspected crimes.
Your professional opinion as a lawyer is that I should not care that my competitor is copying my product marketing texts that I wrote based on my experience with the products, to make it sound like they are as much an expert on those products as I am, thereby unfairly improving their ability to outcompete me by selling at lower prices, due to not having to invest the time necessary to actually be an expert? That is your professional opinion?