Earlier in January, the artificial intelligence model published by the Chinese Deepseek forced Meta’s rapid action. Meta set up four teams to find out how to train Deepseek at such a low cost.
In January, Deepseek, a Chinese start -up company, came to the artificial intelligence market in January. Joel Maalmi
Technology giant Meta brought together four teams to find out how the Chinese Deepseek was able to create his own language model at a very low cost, Fortune says based on the news of The Information.
Earlier in January, Deepseek released its own generative artificial intelligence model R1, which it says is as well as other models, but only a fraction of the cost of competitors.
Deepseek is a Chinese Start Up company that develops a Generative artificial intelligence by Hedge Fund High-flyer.
Open source technology used by Deepseek has put the giant artificial intelligence investments made by US companies in questionable light.
The leader of Meta’s artificial intelligence projects Mathew Oldman has said within the company that Deepseek’s new language model may even be better than developed by meta Llama. According to Mark Zuckerberg, CEO of Meta, Llama’s next version may be released in early 2025.
Two of the four teams set up by Meta are trying to find out how the High-Flyer managed to reduce Deepseek’s cost of training the language model. The aim of the teams is to make use of the information obtained in the development of the meta Llama language model.
One of the other two teams is trying to find out what kind of information Deepseek was used to train a language model. Another team, in turn, is thinking about how Llama could be rebuilt based on the characteristics of Deepseek’s language models.
Zuckerberg announced last Friday that the company would use up to $ 65 billion, or about $ 62 billion for artificial intelligence projects in the coming year.
Zuckerberg’s announcement came just a few days after the Openai underlying ChatgPT told about his collaboration with Softbank, Oracle and other entities sponsored by the White House $ 500 billion, or about $ 475 billion Stargate-The intelligence infrastructure project. The aim of the project is to build dozens of new data centers across the United States.