11 mo. ago • 98%

OpenAI suspends ByteDance’s account after it used GPT to train its own AI model.

www.theverge.com OpenAI suspends ByteDance’s account after it used GPT to train its own AI model.

In today’s issue of Command Line, I reported that ByteDance has been violating the developer license of both Microsoft and OpenAI by using GPT-generated data to train its own, competing model in China. After my report was published, OpenAI spokesperson Niko Felix sent the following statement confirm...

Hacker News @derp.foo OpenAI suspends ByteDance's account after it used GPT to train its own AI model

35 comments

204
- Excellent. I also would have accepted the two Spiderman meme.
  
  29
  
  I thought about that too, but then decided to go with Bender because he's a robot built by an ominous megacorp. Plus he has an attitude that pretty much sums it all up.
  
  11
- Boomers be like
  
  11
  
  "I worked hard for my inheritance and my summer minimum wage job paid for my house and college education"
  
  14
Hahaha. They’re totally okay with stealing from people but hate it when others do the same to them. What a shame!

82
- Reminds me of when Microsoft Bing was stealing Google data, and Microsoft was like "So what?"
  
  https://arstechnica.com/information-technology/2011/02/google-catches-bing-copying-microsoft-says-so-what/
  
  15
Please don't use this software to do things it's capable of

Lmao.

54
I used the AI to TRAIN the AI!

38
- Distilling has been around since forever. It's a legitimate technique that can give you a better model depending on your needs.
  
  OpenAI does it too to improve its models.
  
  26
- People use compilers to compile compilers, or just generally computers to design computer hardware, all the time. It's not so strange.
  
  8
  
  I thought it was almost entirely the point to automate stuff. We have a great tool to automate the automation.
  
  4
- It’s AI’s all the way down!
  
  —See you in City 01.
  
  7
- It's like pupils learning from teachers, with the exception that there's a single teacher in the whole world
  
  6
“Open”AI. This blocking manoeuvre kills progress.

31
- I don't know about that. Training your AI on someone else's AI feels a lot like drinking someone else's piss. I doubt you are going to extract much innovation out of that
  
  10
  
  Can’t be any worse then training it on uninspired social media comments.
  
  17
  
  It works pretty well. You can create a good dataset for a fraction of the effort and price it would have required to do it by hand. The quality is similar. You just have to review each prompt so you don't train your model on bad data.
  
  3
  
  Then they shouldn't interrupt their competitors commiting this mistake right?
  
  3
Well, that's not very open of OpenAI

31
Even the plagiarism machine gets plagiarized nowadays...

27
Not so nice when it happens to you I guess

17
It's as if Sam is just talk and doesn't care about the Open part at all.

14
Oh the irony. Just train it on copyrighted material without a license, like OpenAI did.

11
Boring af the A.I. Wars are.

8
They put the "Ope" in Open.

8
- They put the "N" in Open, which stands for Not Actually Open At All.
  
  12
  
  It stands for Open Platform for Everyone, NOT!
  
  2
It's models all the way down. AI trained on AI trained on AI trained....

5
ClosedAI

4
One of their accounts, anyway...

2
These kinds of articles are the worst, it's all inferred with no qualified actual sources for the claim. And worst of all, it doesn't make sense.

No one uses Ais to train other ais because it doesn't work. It falls apart really fast and is a well-known problem in the ai community.

Bytedance aren't using gpt genersted data to train their own ai because it doesn't work. They may be accidentally including ai generated data in their own training data, but that's just gonna break their own models, so who even cares.

This isn't corporate espionage. This isn't anything but shit journalism.

0

You've viewed 35 comments.