Openai CEO's Sam Altman has neglected the importance of a new artificial intelligence (AI) model released by Chinese startup Deepseek on Thursday, and has made “the wonderful thing of a couple”, but is “very exaggerated.” I mentioned.
DeepSeek's R1 model released last week shakes the American AI industry. This startup claims that it costs only $ 5.6 million to train models with thousands of shortened chips, and questions the investment level required for the development of AI.
“They did some really great things, but as a whole, they are very exaggerated,” said Altoman at an event at Washington DC, “I repeat, but this is a long -lasting model. ”
“I have a sense of what I need to execute a high -cost model,” he continued. “So I don't want to reduce my job. I think this is a great job, but in a sense, it's like Openai's first surprise and new competitors.”
The rise of DeepSeek threatens to reverses the current consensus on AI, assuming that large -scale investment in chips and data centers is needed to build a new AI model. Major AI companies like Openai have recently spent billions of dollars on infrastructure.
The consensus symbol is the Trump administration's new Star gate project, which is trying to invest up to $ 500 billion in the next four years.
Altman said on Thursday that a more large calculation and infrastructure were needed to build a frontier model and to keep up with AI recruitment.
“Costs for the frontier model, very advantage, need training, calculation, test time, calculation, data, and necessary costs,” he said. Using AI of a specific level is much higher than the decrease in prices, as you lower that price. “
He also appeared in Cavalier about the possibility of distillation, which is used to transfer the knowledge of large models to a smaller model.
Openai states that DeepSeek is investigating whether the model is “inadvertently distilled” data. The company's Terms of Use Bar users are building a model that uses distillation to compete.
“People will distill our model,” said Altoman. “Whatever, it's okay.” “” “
Openai CEO advertised DeepSeek's R1 model earlier this week as “impressive”, but was not as worried as a technology sector or some of its surroundings.
“Deepseek's R1 is an impressive model, especially for prices,” says Altman on a social platform X on Monday night. Raise some release. “
Others have a more sensational approach. President Trump calls DeepSeek as the “Morning Call” in the American AI industry, and former Google CEO Eric SC HMIDT may represent the “turning point” of a global AI race for an open source model. I mentioned.
