Jin Lei from Au Fei Temple
量子位 | 公众号 QbitAI
ByteDance has fallen into a large-scale model public opinion storm.
据The Verge报道:
ByteDance has been secretly using OpenAI's technology to develop its own large language model (LLM).
Shortly after the news was revealed, The Verge further said that OpenAI had suspended ByteDance's account.
具体而言,OpenAI发言人Niko Felix发布的声明如下:
While ByteDance uses a small amount of our API, we have suspended their accounts while we investigate further.
If we find that their use is not in accordance with the rules, we will ask them to make the necessary changes or terminate their account.
The "rules" mentioned here refer to a clear provision in OpenAI's terms of service that the model capabilities provided by OpenAI are not allowed to be used to "develop any AI models that compete with its products and services".
It is understood that ByteDance purchased OpenAI access through Microsoft, but Microsoft has also formulated the same policies as OpenAI.
The Verge said it is making further inquiries with Microsoft about whether it will take the same steps as OpenAI to suspend ByteDance's account.
So, what exactly is going on with this plagiarism storm?
Internal documents were exposed
According to The Verge, the evidence came from an internal document from ByteDance, the chat logs of Lark, an overseas version of Feishu.
The document shows that ByteDance relied on OpenAI's API for development at almost every stage of development, including training and evaluating models, in its basic large language model project, code-named Project Seed.
The "Seed Project" was launched about a year ago and is currently developing two products, one is Doubao, which has been launched in China, and the other is a chatbot platform for commercial users, which is currently under development.
Employees involved in the "Seed Project" are said to be well aware of the consequences of over-reliance on OpenAI's APIs, so they began to discuss how to whitewash evidence through "data masking".
So much so that it is not uncommon for employees to reach the maximum access limit of the OpenAI API.
More specifically, ByteDance used OpenAI's technology more in the early stages of the "seed plan".
According to internal documents, ByteDance issued an order about a few months ago to "stop using GPT-generated text at any stage of model development."
However, it was also at this time that ByteDance released its own large language model, Doubao.
But The Verge said that even at this time, ByteDance still did not stop the violations:
ByteDance continues to use the API in a way that violates OpenAI and Microsoft's terms of service, including evaluating the performance of the model behind the bean pack.
And also said that a person who has first-hand information about the internal situation of ByteDance pointed out:
They say they want to make sure everything is legal, but they really just don't want to get caught.
ByteDance has responded
After The Verge published the report, ByteDance spokesperson Jodi Seth responded as follows:
The data generated by GPT was used to annotate models in the early development of the "Seed Project" and was removed from ByteDance's training data around the middle of this year.
ByteDance is licensed by Microsoft to use the GPT API.
We use GPT to support our products in non-Chinese markets, but in Chinese markets, we use our own models to support Doubao.
As for Microsoft, spokesman Frank Shaw said:
Microsoft AI solutions, like Azure OpenAI Service, are part of our limited access framework, meaning that all customers must apply for and be approved by Microsoft.
We also set standards and provide resources to help customers use these technologies responsibly and in compliance with our Terms of Service.
We have processes in place to detect abuse and will stop access to companies that are found to be violating the Code of Conduct.
Qubits also got in touch with ByteDance for the first time, but ByteDance has not made an official response at present.
We will follow up on the progress of this incident in the comment section.
Reference Links:
[1]https://www.theverge.com/2023/12/15/24003151/bytedance-china-openai-microsoft-competitor-llm
[2]https://openai.com/policies/business-terms
— END —
QbitAI · Headline number signed
Follow us and be the first to know about cutting-edge technology trends