The national security risks of selling AI chips to China far outweigh the benefits of spreading US technology around the world, according to Dario Amodei. Anthropic’s founder and CEO pushes back against recent policies that impose sales like a way to integrate the technology of leading US companies, such as Nvidia, into global ecosystems.
“Are we going to sell nuclear weapons to North Korea because that makes Boeing profitable?” asked Amodei while speaking at the World Economic Forum in Davos, Switzerland, today (Jan. 20). “That analogy should clarify how I see this trade-off – that I don’t think it makes sense.”
In recent months, restrictions on AI chip shipments have been exposed under the Trump administration, in part due to the lobbying of Nvidia CEO Jensen Huang. This pushback gives AI leaders little time to understand the technology’s advancements, societal implications and inherent risks of the new technology, according to Amodei. “The reason we can’t [slow down] it is because we have geopolitical enemies who are building the same technology at the same speed,” he explained.
Demis Hassabis, CEO of Google DeepMind, emphasized the need for a flexible approach to the world’s AI challenges while speaking to Amodei at the Davos panel. When it comes to establishing security standards, international cooperation like the US and China is “very necessary,” Hassabis said.
Hassabis added that his concerns extend beyond government to education. He said he has been “constantly amazed” by the way a few economists and professors are examining the effects of AI on issues such as job displacement and wealth distribution.
Getting the public deployment of AI right, Hassabis argued, will require the technology’s evolution to slow down. Achieving that reduction, however, “will require some cooperation.”
The early days of AI’s operational implications
Both Amodei and Hassabis say they are already seeing the impact of AI on the labor market in their companies. Hassabis pointed to a “slowdown” in hiring at Google DeepMind, particularly entry-level roles such as interns.
Amodei, on the other hand, has long warned that AI could cause major disruptions in the workforce. Last year, he said this technology could eliminate 50 percent of all white-collar jobs within five years. He said the change is already happening within Anthropic, and his company is thinking internally about how to manage the change “in a way that makes sense.”
Both leaders believe the job market will eventually adapt, including creating new AI-enabled roles. However, Hassabis stressed that work is more than just a salary. Questions about how meaning and purpose relate to jobs are among those that “keep me awake at night,” he said, adding that the financial consequences of employee disruptions are easier to resolve “than what happens to the human condition, and humanity as a whole.”
Although the CEOs of Anthropic and Google DeepMind largely agree on the political, social, and labor implications of AI, they are at odds over time. Amodei believes that AI can reach Nobel prize-winning capabilities within just a few years. Hassabis, by contrast, puts the odds of human-level AI at 50 percent by the end of the decade.
Still, both agree that neither timeline leaves much room for companies, policymakers, or governments to develop a coherent response to the growing influence of technology. “There is not much time before this comes,” said Hassabis.
