A Chinese AI startup, DeepSeek, has made a big splash in the AI world. Its chatbot has become the most downloaded free app on the U.S. Apple App Store, surpassing OpenAI’s ChatGPT.
DeepSeek’s AI assistant is powered by the company’s open-source models, which they claim can be trained much cheaper and faster than the leading models like GPT-4.
DeepSeek’s R1 reasoning model, released on January 20, is designed to solve complex problems. It reportedly performs similarly to OpenAI’s GPT-4, but at a fraction of the cost.
DeepSeek claims that developing the R1 model cost less than $6 million, while OpenAI’s GPT-4 cost over $100 million to train.
The company also says it only needed 2,000 specialized Nvidia chips to train its models, compared to the 16,000 chips required by other top AI models. These claims have raised eyebrows in the tech world, especially since Nvidia is a key supplier of chips used in AI training.
The news has caused Nvidia’s stock to drop more than 12% in pre-market trading. This has made investors question whether the big investments in AI data centers, like Nvidia’s $100 billion stake in the Stargate Project, are still a good idea.
As DeepSeek’s rise continues, the AI industry is watching closely to see if this new approach to training AI models could change the future of the market.
Leave a comment