Nathan Gardels is the editor-in-chief of Noema Magazine.
The debut of China’s DeepSeek chatbot reaffirms that one of the oldest maxims about human behavior applies to the latest technologies: Necessity is the mother of invention and innovation.
Both geopolitical delusions and stock bubbles burst last week when it became clear that DeepSeek’s generative AI model was on par with the best of the West. More stunning is that it apparently used mostly far less advanced chips and costs mere millions to build compared to the billions American Big Tech is spending for large quantities of the most powerful semiconductors banned for export to China. The ingenuity of DeepSeek’s engineers has been largely attributed to the need to figure out how to work around Western restrictions.
This paradox of Western efforts to stymie China’s tech development actually stimulating it is not new. As Zheng Yongnian, a
Shenzhen-based scholar closely listened to by the Chinese leadership wrote in Noema in 2018, the tech restrictions already then imposed may slow the pace of China’s leap forward but will “not be able to stem its high-tech catch-up. China’s enormous reserves of state capital, its substantial pool of ready talent and its huge market will continue to drive it forward.”
He continued: “Indeed, the United States may be shooting itself in the foot. U.S. trade hawks fail to grasp the record of contemporary history, which demonstrates that technology bans imposed by the West are more often than not counterproductive when directed at a capable country.
For example, China was compelled to go solo on projects such as its manned spaceflight program after being cut off by rigorous exclusion policies imposed by the West. Also, strict technology bans by America have forced China to lay out huge national investments that led to its exceptional performance in building supercomputers that surpass those of the United States. And in the aftermath of the 1989 Tiananmen Square movement, the full-scale embargo on military technology imposed on China by the West propelled and motivated China to develop independent military weaponry systems.”
At the same time, advanced generative AI models like DeepSeek, which piggyback on Western training data, pose their own dangers within China. What do Chinese authorities do when these models generate something that’s not permitted in an otherwise tightly controlled information space? Will ever-smarter chatbots figure out how to circumvent censors?
Open Source Vs. Closed Models
As an inexpensive open-source model, DeepSeek will be widely available and accessible to anyone anywhere in the world to build upon. That will no doubt spur new waves of innovation across the planet that will perennially challenge the dominance of the closed, proprietary models of Big Tech that drive AI development in the U.S. Meta’s Llama is the only open-source exception among these large U.S.-based players.
Whether for good or for ill, open-source software married to the resourcefulness of collective intelligence will know few bounds.
For former Google CEO Eric Schmidt, this is why the arrival of DeepSeek is a “turning point” in global AI competition that must change the way Silicon Valley innovates.
“The United States already has the best closed models in the world. To remain competitive, we must also support the development of a vibrant open-source ecosystem,” Schmidt wrote with Dhaval Adjodah in The Washington Post. “The race between open- and closed-source AI, as well as between the United States and China, does not yet have a clear winner. But there is clearly mounting pressure on America’s Big Tech players if DeepSeek can compete with them using far fewer resources.
It is unlikely that American frontier model companies will change their business models anytime soon, nor is it immediately clear that they should,” the authors continued. “Open and closed competition will most likely find a natural equilibrium, with a range of different offerings and price points for different users.”
The Downside Of Open Source
When Schmidt alludes to it not being a good idea for closed frontier AI model development to entirely open up, he is referring to the downsides of open source.
In a conversation with Noema last May he laid out his concerns.
“It’s always useful to remind the techno-optimists in my world that there are evil people. And they will use your tools to hurt people. … All technology is dual use. All of these inventions can be misused, and it’s important for the inventors to be honest about that,” Schmidt warned, referring to all manner of threats from bioweapons and automated warfare to deep-fake disinformation and facial recognition for repressive purposes.
“Whether for good or for ill, open-source software married to the resourcefulness of collective intelligence will know few bounds.”
Schmidt went on: “In open-source and open-weights models the source code and the weights in models [the numbers used to determine the strength of different connections] are released to the public. Those immediately go throughout the world, and who do they go to? They go to China, of course, they go to Russia, they go to Iran. They go to Belarus and North Korea. When I was most recently in China, essentially all of the work I saw started with open-source models from the West and was then amplified.”
This, indeed, is the case with DeepSeek.
While Schmidt expects that the leading AI firms in the West putting billions into proprietary AI models “will eventually be tightly regulated as they move up the capability ladder [toward artificial general intelligence],” he worries “the rest will not.”
When asked about the wisdom of trying to stop China’s advance by banning the export of chips, he replied, “I’m much more concerned about the proliferation of open-source. And I’m sure the Chinese share the same concern about how it can be misused against their government as well as ours.
We need to make sure that open-source models are made safe with guardrails in the first place through what we call ‘reinforcement learning from human feedback’ (RLHF) that is fine-tuned so those guardrails cannot be ‘backed out’ by evil people. It has to not be easy to make open-source models unsafe once they have been made safe.”
Just as necessity is the mother of invention, the dual-use character of intelligent technology also means that innovation is the mother of its own necessary regulation. That must be especially so when fully capable chain-of-thought reasoning empowers AIs to acquire their own agency.
“For most of history,” DeepMind’s Mustafa Suleyman has said, “the challenge of technology lay in creating and unleashing its power. That has now flipped; the challenge of technology today is about containing its unleashed power, ensuring that it continues to serve us and our planet.”
