OpenAI: The Chinese company DeepSeek relies on Distillation to benefit from ChatGPT and American models.. What does that mean?

OpenAI, the AI company and developer of ChatGPT, has accused the Chinese AI company DeepSeek of "ongoing efforts to freely exploit the capabilities developed by OpenAI and other leading American labs."

How are they doing this?

OpenAI stated in an official warning to American lawmakers that DeepSeek employees:

- Develop methods to bypass access restrictions to OpenAI models.
- Access models through third-party networks that hide their source.
- Develop code to obtain outputs from American models for the purpose of Distillation.

What is Distillation technology?

It is a process used in training AI models, where a large and powerful model (the "teacher" model) evaluates the quality of responses generated by a new and smaller model (the "student" model).

In this way, expertise and knowledge are transferred from the advanced model to the new model, enabling it to learn effectively and at a much lower cost.

Why does this issue concern Washington?

DeepSeek-V3 and DeepSeek-R1 models have caused a stir in the markets, as they are considered strong competitors to the best American models.

OpenAI believes that large Chinese models "skip important steps in training and deploying new models safely," which threatens innovation and security.

The United States fears that China will catch up in the global AI race despite the restrictions imposed on it.

$AI ‌$AIA
$AIC