Meta’s Chief AI scientist shares his opinion about DeepSeek’s success

Meta’s Chief AI scientist Yann LeCun has given his assessment about the success that DeepSeek is enjoying in the artificial intelligence industry. According to LeCun, the biggest point to note in its rise is its vision to keep AI models open source so that everybody can benefit from it. The Chinese artificial intelligence company recently launched its DeepSeek-R1, which features better reasoning and problem-solving capacities, rivaling industry leaders like OpenAI.

The AI, launched on January 20, possesses remarkable features, with tests showcasing its exceptional abilities. Aside from that, DeepSeek provides training parameters under an MIT license, allowing researchers globally to study and enhance the program.

LeCun noted that the company’s success story is not about putting China ahead of the United States in the AI industry, but making sure open source models are still ahead. It’s not that China’s AI is “surpassing the US,” but rather that “open source models are surpassing proprietary ones,” LeCun said.

Meta’s Chief AI scientist discusses DeepSeek’s success

DeepSeek shares the same attribute with the Llama model created by Meta, with both being open source. Their rival, OpenAI, which was originally created as an open-source AI firm, has since tried to shift its mission to create technology that will benefit all humanity. The move has caused a lot of legal tussle, with Elon Musk, one of its former executives and contributors asking the court to block its transition to a for-profit.

According to Meta’s Chief AI scientist, DeepSeek has enjoyed open source and open research, noting that it was the idea that some people made that they chose to build on. “They came up with new ideas and built them on top of other people’s work. Because their work is published and open source, everyone can profit from it,” LeCun said. “That is the power of open research and open source.”

During the unveiling of DeepSeek R1, the company mentioned that the system demonstrates remarkable reasoning capacity, something that has reportedly been put to the test. The company noted that it is trying to push the boundaries of open-source AI. In its paper, the company also highlighted that it made the system because it wanted to explore the potential of artificial intelligence to develop its reasoning abilities without the use of supervised data.

According to Professor Ethan Mollick at Wharton, the responses gotten from the DeepSeek R1 is like a human thinking out loud. Meanwhile, there is the issue of the R1 providing issues to AI companies in the United States, seeing that both China and the US have been rivals for a while. With Trump in the White House again, the rivalry is expected to heat up. However, the rivalry is not the main concern, as the AI model shows that China will very much challenge the top AI companies in terms of development.

Silicon Valley surprised by DeepSeek announcement

The launch of the DeepSeek R1 model took Silicon Valley by surprise, taking center stage in tech discussions in a week that featured various events including Trump’s inauguration and his busy first few days in office, the World Economic Forum that took place in Davos, Switzerland, and TikTok cutting and restoring service to users in the United States.

Days after DeepSeek unveiled its model, Meta CEO Mark Zuckerberg announced plans to spend over $60 billion in 2025, as the firm prepares to invest big in artificial intelligence. The Meta boss has been a very outspoken proponent of open-source models, even going as far as backing his rival, Elon Musk, in his quest to block OpenAI’s for-profit move.

In September 2024, the Meta CEO mentioned he wanted to build the next generation of open platforms. Zuckerberg noted that he hopes to achieve it in the next 10-15 years, stressing that open platforms must win. “I think that’s going to lead to a much more vibrant tech industry,” he added. Making a technology open source has been hailed as the way to go, since it allows anyone to modify and redistribute the codes, this way the technology develops rapidly in a democratic setting.

However, proponents of a closed source system have also argued the merits of the models, noting that they are more secure because the codes are kept private. In his case, OpenAI CEO Sam Altman has mentioned that taking the closed-source approach with the company provides them an easier way to hit the safety threshold. However, he mentioned in the AMA on Reddit that in the future, he might try to open-source more stuff.

Cryptopolitan Academy: FREE Web3 Resume Cheat Sheet - a href="https://www.cryptopolitan.com/ultimate-web3-resume-cheatsheet?utm_source=cryptopolitan&utm_medium=banner&utm_campaign=web3cs-1" target="_blank">Download Now