>"OpenAI, backed by Microsoft, suspends ByteDance's account for reportedly using its technology to develop a large language model, breaching terms."
>"This move followed the revelation that ByteDance had covertly employed OpenAI's technology to develop its own AI model, codenamed "Project Seed." "
>" According to The Verge, this practice is generally considered a faux pas in the AI world. It's also in direct violation of OpenAI's terms of service, which state that its model output can't be used "to develop any artificial intelligence models that compete with our products and services." OpenAI is backed by Microsoft, which ByteDance is buying its OpenAI access through, and has the same policy."
https://www.ndtv.com/world-news/openai-suspends-bytedances-account-over-use-of-gpt-for-training-its-own-ai-model-4700732/amp/1
https://www.theverge.com/2023/12/15/24003151/bytedance-china-openai-microsoft-competitor-llm
Yikes.... Are the feds really THAT scared of China catching up in the AI Arms race?
Shopping Cart Returner Shirt $21.68 |
Tip Your Landlord Shirt $21.68 |
Shopping Cart Returner Shirt $21.68 |
>ByteDance had covertly employed OpenAI's technology to develop its own AI model
how?
Read the article. It sounds like they used the API to ask chat GPT a bunch of questions regarding training their own AI. Likely be a heavy use of the paywald GPT-4 model and used some special tools that took advantage of the API.
>feed an existing LLM some prompts
>save the prompt+output as training data
>use it to train a model that imitates the other LLM
It's lower quality training data than preparing your own, since there's a good chance it will amplify any defects/biases from the original LLM.
But it could, under some circumstances, be much cheaper than preparing your own training data.
It's also for data collection. They say 'x', we say 'y', why is there a difference.
LLMs hallucinate faster with LLM conversations feed into the training data.
You can get an LLM close in performance and FAR more lightweight by following what these guys did
https://crfm.stanford.edu/2023/03/13/alpaca.html
Pretty sure it was a Microsoft research paper that convinced everyone to do this.
link it, Black person/pajeet
https://www.microsoft.com/en-us/research/publication/orca-progressive-learning-from-complex-explanation-traces-of-gpt-4/
https://medium.com/@saikrishna.v1970/the-next-big-wave-in-ai-unveiling-orca-microsofts-game-changing-innovation-8d0fb34d2238
>Project Sneed
>suspends their account
So just make a new account with a different employee's credentials? Why do people act like "we banned your account" is a big deal? Don't people know you can ban evade?
In most cases it's no big deal. In this case it might be foreshadowing a lawsuit.
>Why do people act like "we banned your account" is a big deal?
social conditioning
Why didn't they just use a vpn? morons
>how could we have possibly known the @bytedance.com account with $10,000,000 in spend was abusing the API?
>Are the feds really THAT scared of China catching up in the AI Arms race?
Wasn't China supposed to be far ahead in AI?
project sneed