According to Financial Times reports, Openai has discovered that DeepSeek has illegally used its own models to train open source competitors.
Chatgpt manufacturers argue that DeepSeek has used a method called “distillation.” This allows small models to duplicate the performance of a large model with few computing resources. Openai claims that it constitutes the unauthorized use of the technology and causes concerns about potential intellectual property theft.
Openai says that Deepseek in China has evidence of training open source AI using that model.
The San Francisco -based AI company told Financial Times that it had gathered evidence that DeepSeek was using distillation to improve the AI model. This method, which is commonly used in machine learning, allows developers to use output from more advanced systems to improve smaller models.
“Openai is used by Chinese artificial intelligence, DeepSeek, to train our own open source competitors using US companies’ unique models as concerns are increasingly concerned about the potential violations of intellectual property. Financial Times reports that you have found evidence that you have done.
Financial Times continued:
“Based in San Francisco [ChatGPT] Maker told Financial Times that he had seen evidence of distillation from DeepSeek. This method is used to improve the performance of a smaller model by using output from a larger and more competent model, and at a much lower cost in a specific task. I will be able to achieve it. “
In response, Openai stated that it has strengthened the cooperation with the US government to protect American AI technology, especially from foreign threats from China. The company emphasized the importance of securing intellectual property and maintaining advanced functions built into the model.
DeepSeek’s AI has already shaken the market and has contributed to a large -scale $ 700 billion wipeout in high -tech stocks. The company’s highly expensive models bring direct challenges to US companies that control the AI space.
Openai CEO’s Sam Altman pointed out that DeepSeek models were recognized as a powerful competitor and needed more computing power to maintain the edge. Former President Donald Trump stressed the situation and emphasized the importance of keeping the United States at the forefront of technology.
Despite the reform, some analysts are looking at the advantages. They believe that improving the efficiency of DeepSeek can further upload the adoption of AI.
According to multiple reports, Deepseek V3 has exceeded the main models such as LLAMA 3.1 and GPT-4O on major benchmarks, including CodeForces competitive coding issues. This project was completed with a budget of only $ 5.5 million. This is completely contrasting with hundreds of millions of rivals spent. This breakthrough is challenging the concept that cutting -edge AI development requires enormous financial investment.
On the other hand, there are many questions. If the Openai model is a closed source, how did you access DeepSeek? Another question is that if DeepSeek steals Openai’s unique model, why is Openai better than Deepseek?
Source link
