The high-tech bosses agree that Deepseek’s risk pose will remain limited to Openai for now.
Bloomberg | Bloomberg | Getty Images
The technological advances exhibited by the Deepseek in China’s artificial intelligence lab show that the game is on when it comes to the US-Sino competition in AI, a Top Tech executive told CNBC.
In a series of interviews at the French Artificial Intelligence Action Summit, leaders of several major tech companies told CNBC that the emergence of Deepseek shows China cannot count as a serious player when it comes to AI innovation. Ta.
Last month, Deepseek shocked the global market with a technical paper that said one of its new AI models was created with a total training cost of less than $6 million. Labs for Openai and humanity.
Openai’s Chief Global Affairs Officer, Chris Lehane, has confirmed that Deepseek’s advanced low-cost model “has a very realistic competition between US-led small D-democratic AI and CCP. He said he was there. [Chinese Communist Party] A dictatorial and authoritarian AI led by China. ”
Many Deepseek critics point to obvious censorship by the model when it comes to sensitive topics. For example, when asked about the 1989 Tiananmen Square Massacre, the AI assistant app from Deepseek responds: “Sorry, that’s beyond my current scope.
“There are two countries in the world that can build this on a large scale,” Lehaan told CNBC’s Arjun Kharpal on the sidelines of the Paris AI Summit on Monday. “Imagine that there were only two countries that could build electricity on a large scale, and that’s like you need to think about it.”
“What Deepsee really strengthens and reaffirms is that there are very realistic stakes and very realistic competition,” added Lehane.
Still, despite Tech Bosses showing that Deepseek’s breakthrough is even further in a global AI race than previously thought, the threat posing for Openai remains limited for now. I almost agreed to something.
“The game is on”
Deepseek is its new R1 model, An The open source inference model could rival the performance of Openai’s own similar O1 model.
This has led experts to question the general wisdom of the West in the past few years. In other words, China is behind the US when it comes to AI development. Processing unit or GPU.
GPUs are necessary for training and execution as AI applications are superior in parallel processing. This means that multiple calculations can be performed simultaneously.
Reid Hoffman, co-founder of LinkedIn and partners at venture capital firm GreyLock Partners, told CNBC Monday that Deepseek’s new model is “a big deal to show that the game is on.”
“The competition is ongoing with China,” Hoffman said, adding that Deepseek’s R1 is a “reliable, practical model.”
Abishur Prakash, founder of strategic advisory firm The Geopolitical Business, told CNBC that Deepseek shows limited understanding of western China.
“The American assumptions as a global technology captain is no longer an acceptable belief,” Prakash told CNBC in a telephone interview.
“That’s the new status quo right now, the space between the US and China is getting narrower almost overnight, but not overnight, that’s been a long-standing progress,” Prakash said. Ta.
“If there’s one takeaway in the West, their understanding of China is very limited, and we don’t know what’s coming next,” he added.
There is no meaningful threat to us – yet
Still, the executives of major AI companies are not convinced that Deepseek is poses any meaningful risk to AI lab businesses like Openai and humanity.
While the entire experts agree that Deepseek’s AI advancements are impressive, questions have been raised about startup claims about costs.
A report from Semiconductor research firm Semianalysis estimates last month that Deepseek’s hardware spending is “much higher” than $500 million in the company’s history. When contacted by CNBC, Deepseek could not immediately comment.
According to the report, the costs associated with Deepseek’s research and development costs and ownership are important, and “substantial amounts of calculations” are required to generate “synthetic data” for the model to train. I understand.
Some engineers believe DeepSeek could have achieved such a high level of performance by training the model on a larger US AI system.
Known as “distillation,” this technique involves assessing the quality of the answers generated by the new model into a more powerful AI model.
The claim that Openai itself is implied, reviewing a report last month on CNBC that Deepseek may have developed an AI model called “distillation” using output data from the model “inappropriate”. It’s an assertion that there is.
“Most of the market is afraid of those around me. [DeepSeek] In fact, Hoffman told CNBC.
“I think the short answer everyone should take is Game-on, but the larger models are still really important,” he added.
Victor Riparbelli, CEO of AI Video Platform Synthesia, told CNBC that Deepseek “challenged the paradigm that brute force scaling is the only way to build a better model, but the idea that companies will suddenly change large quantities. The AI workload is misplaced.
“If you look at the users of these technologies, all of the workflows, I think that looking back in three months, 0.01% of them will move from Openai and humanity to Deepseek,” Riparbelli said.
Meredith Whitaker, president of Signal Foundation, said Deepseek’s development hasn’t moved much of a needle for the industry as market momentum still supports a larger AI model. Signal Foundation is a non-profit organization that supports encrypted messaging app signals.
“This does not disrupt the concentration of force or geopolitical balance at this stage,” Whitaker told CNBC. “I think we have to keep an eye on the ball there and realize that it’s this ‘big more’ paradigm and that it’s not diminished with historically efficient benefits.
