The tech bosses largely agree that the risks posed by DeepSeek to Openai are still limited at present.
Bloomberg|Bloomberg|Getty Images
Top tech executives told CNBC that technological advances demonstrated by China’s artificial intelligence lab DeepSeek show that in-game competition is in progress.
In a series of interviews at the French AI Action Summit, leaders of several major tech companies told CNBC that the emergence of DeepSeek shows that China cannot be seen as a serious player when it comes to AI innovation.
Last month, DeepSeek shocked labs such as Openai and Anthropic in global markets with a technical paper.
Chris Lehane, chief global affairs officer at OpenAI, told CNBC that DeepSeek’s advanced, low-cost model confirms there is a “very real competition between US-led, small D democratic AI and CCP (Chinese Communist Party) China-led autocratic, authoritarian AI.”
Many critics about DeepSeek have pointed out that the model has been clearly reviewed on sensitive topics. For example, when asked about the Tiananmen Square Massacre in 1989, DeepSeek’s AI Assistant app responded: “Sorry, this is beyond my current scope. Let’s talk about other things.”
“There are two countries in the world that can build this on a large scale,” Lehane told CNBC’s Arjun Kharpal on Monday, on a sidelines at the Paris AI Summit. “Imagine if there were only two countries in the world that could build on a large scale.” Electricity. That’s how you have to consider it.”
“For us, the real strengthening and reiterating is that there is very real competition and has very real shares,” Lehane added.
Still, tech bosses agree largely, despite DeepSeek’s breakthrough suggesting that the threat posed by China to Openai is still limited in scope among the global AI race.
“Game Starts”
DeepSeek says it is the new R1 model, a Open source reasoning modelable to rival the performance of OpenAI’s own O1 models – only use cheaper, less energy-intensive processes.
This has led experts to question the general wisdom of the West over the past few years, which is China lagging behind AI development in terms of AI development, as export restrictions make it harder for companies in the country to master more advanced NVIDIA graphics processing units or GPUs.
GPUs are required for training and running AI applications because they perform well in parallel processing, which means they can perform multiple computations simultaneously.
Reid Hoffman, co-founder of LinkedIn and partner at venture capital firm Greylock Partners, told CNBC on Monday that DeepSeek’s new model “is important, which shows that the game has begun.”
“The game with China is being played against China.
Abishur Prakash, founder of the geopolitical business of strategic consulting firm, told CNBC that DeepSeek shows that Western understanding of China remains limited.
“The position of the United States as the world’s technical captain is no longer an acceptable belief,” Prakash told CNBC in a telephone interview.
“This is the new status quo, where the space between the United States and China has shrunk almost overnight – but it hasn’t shrunk overnight, it’s been years of progress,” Prakash said.
He added: “If there is one thing the West gains, it is that their understanding of China is very limited – we don’t know what will happen next.”
There is no meaningful threat to us, but
Still, leading AI executives don’t believe DeepSeek poses any meaningful risk to the businesses and numerous businesses in Openai and Anthropic’s AI labs.
While it is impressive to agree with DeepSeek’s AI advancements overall, questions are raised about the startup’s claims on costs.
A report by Semiconductor Research firm Semi-analytics estimates that DeepSeek’s hardware spending is “higher than $500 million in the company’s history.” When CNBC contacted DeepSeek, there was no immediate comment.
The report found that DeepSeek’s R&D costs and expenses related to ownership are important, and that “synthetic data” trained for the model will require “a considerable amount of computation.”
Some technicians believe that DeepSeek may be able to achieve such high performance by training its models on larger U.S. AI systems.
The technology, known as “distillation”, involves making more powerful AI models evaluate the quality of answers generated by new models.
Claims Openai itself has mentioned that it told CNBC in a statement last month that it is reviewing DeepSeek’s report that DeepSeek may have used output data from its model “inappropriately” to develop its AI model, which is a A method called “distillation”.
“Most market fears (DeepSeek) are actually misplaced,” Hoffman told CNBC. “It still requires large models – it’s extracted from large models.”
He added: “I think the short answer that everyone should take is: the game – but the large models are still very important.”
AI Video Platform Synthesia CEO Victor Riparbelli told CNBC that while DeepSeek challenges “Brute Force scaling is the only way to build better and better models,” the company will suddenly shift a lot of scope and AI workloads are misled.
“I still think that when you look at the users of these technologies, all the workflows, I think when we look back three months later, I think 0.01% of them will be moved from Openai and Anthropic to DeepSeek,” Riparbelli said .
Meredith Whitaker, president of the Signal Foundation, said DeepSeek’s development has not brought much momentum to the industry, as market momentum still widely supports larger AI models. Signal Foundation is a nonprofit organization that supports encrypted messaging application signals.
“This doesn’t undermine the attention of strength concentration or geopolitical balance at this stage,” Whitaker told CNBC. “I think we have to focus on the ball there and realize that it’s really the ‘bigger’ paradigm, and Not reduced by historically increasing efficiency, which prompts this concentration.”
