Mark Zuckerberg, CEO of Meta Platforms Inc., arrives at the Meta Connect event at Menlo Park in California, September 25, 2024.
David Paul Morris | Bloomberg | Getty Images
Target Executive director Mark Zuckerberg said on Wednesday that it was too early to say what the impact of Deepseek’s breakthrough will have a model on his company and the wider technology industry.
Still, Zuckerberg poured cold water on the idea that the target of total consumption would soon fall when Deepseek allegedly proved that top -notch AI requires far less money and a computer resource than before.
“It is probably too early to have a strong opinion on what it means to the trajectory on the infrastructure and the captain,” Zuckerberg said. “There is a bunch of trends that happen here at once.”
AND Technology markets were religion this week over the AI model created by Deepseek, and that is Ai laboratory based in Hangzhou Associated with the Chinese Hedge Fund High Flights. Deepseek claims it only took two months and less than $ 6 million to develop R1 Great Language Model. And news caused by Nvidia shares tank 17% on Monday due to implications that companies do not need so many graphic units to process chips for their own artificial intelligence projects.
Zuckerberg talked about Deepsek during a call with analysts on Wednesday as part of Meta’s The earnings of the fourth quarter. The company reported that sales increased by 21% compared to the year to $ 48.39 billion, on the eve of the Wall Street expectations.
Meta’s shares on Monday increased by almost 2%, indicating that some investors believe that Deepseek’s economy could mean that the company of social media could create powerful AI systems for less money. Meta announced last week plans invest between $ 60 and $ 65 billion This year as part of his AI pressure. The company said on Wednesday that its total costs 2025 will reach between $ 114 billion and $ 119 billion.
The target still digests some of the Deepseek features and that his team hopes that he will eventually apply some of these progress for his AI projects, Zuckerberg said.
“It’s part of nature as it works, whether it’s a Chinese competitor or not,” Zuckerberg said, calling for how the progress of one company ends in favor of others.
Although it is possible that the companies will eventually need less computer resources to train these powerful models, Zuckerberg said that he would have a ton of server power may be needed when the software actually triggers and performs actions, resulting in “higher levels of intelligence and higher quality of service.”
“It will be expensive for us to serve all these people because we serve a lot of people,” said Zuckerberg, whose company said she had 3.35 billion active people in the fourth quarter.
The target will continue to learn from technologies like Deepsek and let out their own open source and free Llam AI models that can also push the industry, Zuckerberg said.
“I still think the investment in Capex and Infra will be a strategic advantage over time,” he said. “It is possible that at some point we will learn differently, but I just think it is too early to call it, and in this point I would bet that the possibility of building this kind of infrastructure would be the main advantage.”
Zuckerberg also said that Deepseek’s phenomenon confirms the target of dedication to the open code of AI, Zuckerberg said.
“There will be a global open code standard,” he said. “For our national advantage, it is important that it is an American standard.”
Watch:: Meta beats on top and lower lines.
