A one-year-old Chinese AI startup, Deepseek, knocked the air out of OpenAI with its latest R1 model. Deepseek is cheaper, open-source, and outperforms OpenAI Chatbot in many categories. It did this with a limited budget of $6 million and strict trade restrictions on powerful chips from the USA.
Instead of fighting back with better technology, OpenAI has resorted to slander. Sam Altman ran out of his office screaming, “They copied us.” Microsoft is in on the fun too. Microsoft is a big investor in OpenAI Chatbot and hosts majority of OpenAI’s data on its servers.
Even Donald Trump’s AI czar, David Sacks, appeared on Fox News to complain about how Deepseek is using OpenAI chatbot to make better models through a process known as “knowledge distillation.”
The Truth About Deepseek and Knowledge Distillation
OpenAI claims to have found “solid evidence” linking Deepseek’s success to knowledge distillation. Distillation is a common technique for training AI and LLM models by building upon larger, more capable models.
Knowledge distillation is an effective way to train smaller models like Deepseek at a fraction of the cost, whereas it took more than $100 million for OpenAI Chatbot to train their GPT models. While other developers and companies can use ChatGPT AI to create their models, using knowledge distillation to create a rival model is a violation of OpenAI’s IP rights.
However, OpenAI has yet to provide any evidence or file any formal legal actions. Microsoft also supported OpenAI’s claim by stating that they have detected large amounts of data being exfiltrated through OpenAI’s API. They traced the location back to China and claimed that these accounts could be linked to Deepseek.
OpenAI Hypocrisy Over Copyright Exploitation
OpenAI’s complaint over copyright infringement is ironic. OpenAI Chatbot was able to make massive leaps by consuming everything on the Internet without the users’ consent. This written data contains information from major publications like The Associated Press, The New York Times, and Axel Springer.
OpenAI has already made deals with the Associated Press, Axel Springer, the Financial Times, and NewsCorp. However, the New York Post has filed a lawsuit against OpenAI for using its content without permission. The lawsuit could result in OpenAI paying the New York Times a minimum of $7.5 billion.
OpenAI is working hard to maintain its dominance over the AI market, but Deepseek is not its only problem. Amidst fierce domestic competition, Chinese AI giants like Alibaba’s Qwen and Kimi K1.5 are surpassing many of OpenAI’s models.
Read Next: How Deepseek Became the Ultimate Disruption Silicon Valley Fears