In a policy document published on Wednesday, former CEO of Google Eric Schmidt, CEO of AI CEO Alexandr Wang and the AI Security Director Center, Dan Hendrycks, said the US should not follow a manhattan press to develop AI systems “
The paper, entitled “Superstitious strategy“He argues that an aggressive offer by the US to exclusively control over -linking AI systems could cause strong retaliation from China, possibly in the form of cyberspace, which could destabilize international relations.
“[A] Manhattan’s work [for AGI] He assumes that opponents will agree on a constant imbalance or strikingly instead of moving to prevent it, “the co-authors write.” What begins as a push for a Superweapon and global risks of control that causes hostile countermeasures and escalating tensions itself. “
A co-author of three extremely important figures in America’s AI industry, the document comes just a few months after a US Committee of Congress proposed an “Manhattan Project-Style” attempt to fund the Agi Development, which was formed after the 1940-year-old US FROM program. recently stated that the US is “The start of a new Manhattan work“At AI while standing in front of a supercomputer location alongside co -founder Openai Greg Brockman.
The Superintelligence strategy document challenges the idea supported by several US leaders of politics and industry in recent months that a program backed by the AGI government is the best way to compete in China.
In the opinion of Schmidt, Wang and Hendrycks, the US is in something that is not different Mutual assurance of destruction. In the same way that world powers do not seek monopolies over nuclear weapons-which could cause a precautionary strike from a Schmidt opponent and co-authors argue that the US must be careful to fight for the dominance of the extremely powerful AI systems.
While AI systems in nuclear weapons may sound extreme, world leaders are already regarding AI as a top military advantage. Already, the Pentagon says AI helps to accelerate the chain of killing the army.
Schmidt et al. They introduce an idea that they call mutually assured AI dysfunction, in which governments could actively deactivate the threatening works of AI and not wait for opponents to arm AGI.
Schmidt, Wang and Hendrycks suggest to the US to shift its focus from “Winning the Race to Superintelligence” to developing methods that Prevent other countries From the creation of Superintelligent AI. Co-authors argue that the government must ‘expand [its] Arsenal of Cyberettacks to disable threatening AI projects “controlled by other nations, as well as limit opponents’ access to advanced AI brands and open source models.
Co-authors identify a partition that has played in the AI policy world. There are “doomers”, who believe that the devastating effects of AI development are a preliminary conclusion and support countries that slow down the progress of AI. On the other hand, there are the “ostriches”, who believe that nations should accelerate the development of AI and essentially simply hope that everyone will work.
The document proposes a third way: a measured approach to the development of AGI that prioritizes defense strategies.
This strategy is particularly remarkable by Schmidt, who has previously shouted at the need for the US to compete aggressively with China in developing AI advanced systems. Just a few months ago, Schmidt released an op-Ed saying that Deepseek marked a turning point in the AI race of America with China.
Trump’s administration seems dead to promote America’s growth. However, as co-authors note, America’s decisions around AGI do not exist in a vacuum.
As people are watching America pushing the AI threshold, Schmidt and co-authors suggest that it may be more prudent to follow a defensive approach.
