Competition in the tech world has always been fierce—artificial intelligence is no different. New players come up with promising breakthroughs that challenge standing giants, and one such player is China’s DeepSeek. Just a week after its launch, DeepSeek took over the tech stock market and even became the top downloaded app on the App Store. However, recent developments have led many to believe that they might have secretly trained their AI using OpenAI’s ChatGPT.
What is DeepSeek and Why the Headlines?
DeepSeek is the AI startup behind the open-source large language model DeepSeek R1. While the company initially received praise for its innovation, reports soon came up that R1 sometimes identified itself as ChatGPT, a feat that had AI enthusiasts alleging that DeepSeek may have used OpenAI’s ChatGPT outputs during its training process.
Knowledge Distillation: The Alleged Method Used by DeepSeek for AI Training.
The basis of these allegations is “knowledge distillation,” a technique where a smaller or newer AI model (the student) learns from a more advanced model (the teacher), to make AI training more efficient. Thus, the student model achieves high performance without the massive computational resources required for training from scratch.
If DeepSeek actually trained its R1 model using responses generated by ChatGPT, this would mean that its AI effectively “Learned” from OpenAI’s propriety model, something I doubt they would be very happy about.
That said, knowledge distillation itself is a common AI training practice, however, the ethical concerns come when it involves data siphoning from another company’s AI without permission, seeing as these two are rivals in the market.
True to point, the AI industry was quick to react, and tech experts and AI researchers began investigating the similarities between DeepSeek R1 and ChatGPT. Some noticed that R1’s responses sometimes came out similar to those of ChatGPT, thus the speculations that OpenAI’s model may have in fact played a significant role in R1’s training.
Following the revelations, some U.S. tech companies saw their stock values dip, with analysts citing concerns over the increasing competitiveness of China-based AI firms. If DeepSeek has indeed managed to develop an AI model rivaling ChatGPT by leveraging OpenAI’s technology, it could reshape AI development, and the AI business model itself globally.
These debates surrounding DeepSeek raise bigger questions about AI ethics and intellectual property rights. Currently, AI companies protect their models through strict licensing agreements, but enforcement remains a challenge. While OpenAI has not yet issued an official statement on the matter, industry observers believe this incident could prompt stricter regulations on AI training data and model sharing.
For now, DeepSeek’s R1 remains a powerful new contender in the AI space. But then again, did it truly innovate, or did it take a shortcut by borrowing from OpenAI’s work?
OpenAI says it has evidence China’s DeepSeek used its model to train competitor – Finacial Times
