Openai reportedly claims that DeepSeek may have distilled its artificial intelligence (AI) model into building an R1 model. According to the report, the San Francisco-based AI company said it has evidence that some users are using the output of their AI models for competitors, which is suspected to be DeepSeek. It is worth noting that the Chinese company released the open source DeepSeek-R1 AI model last week and hosted on Github and Hugging Face. The inference-focused model exceeds the functionality of Chatgpt-Maker’s O1 AI model in several benchmarks.
Openai says it has evidence of foul
According to a report from the Financial Times, Openai claims its proprietary AI model is used to train DeepSeek’s models. The company told the publication that it has seen evidence of distillation from multiple accounts using the OpenAI Application Programming Interface (API). AI companies and their cloud partner Microsoft investigated the issue and blocked their access.
“We know that companies (Chinese) and others have been working hard to thrive in leading the U.S. AI companies model,” Openai said in a statement to the Financial Times. The Chatgpt-MM-Menaker also stressed that it is currently being Work closely with the U.S. government to protect its border model from competitors and rivals.
It is worth noting that AI model distillation is a technology used to transfer knowledge from large models to smaller, more efficient models. The purpose here is to bring smaller models before par or larger models while reducing computational requirements. It is worth noting that OpenAI’s GPT-4 has about 1.8 trillion parameters, while DeepSeek-R1 has 1.5 billion parameters, which will be suitable for description.
When a company is creating a more efficient version of its internal model, knowledge transfer is often the case where smaller models are trained using the relevant dataset of larger models. For example, Meta created several encoding-centric Llama models using the Llama 3 AI model.
However, this is impossible when competitors who cannot access the dataset of proprietary models want to distil out the model. If OpenAI’s allegations are correct, it can be done by adding prompts to its API to produce a lot of output. This natural language data is then converted into code and fed to the base model.
It is worth noting that OpenAI has not made any public statement about this. Recently, the company’s CEO Sam Altman praised DeepSeek for creating such advanced AI models and increasing competition in the AI field.
