Dr. Michael Hochberg & Marcus Gomez
The U.S. Department of Commerce recently sought to enhance export controls on artificial intelligence (AI) hardware and added several People’s Republic of China companies to the entity list.[1] This is a step in the right direction, aimed at containing China’s ability to weaponize AI. But it’s not enough.
Progress in AI models is moving at an astounding pace: The most advanced models are now producing glimmers of what may be considered strategic action and self-protection.[2] These models are like nuclear weapons were in the 1950s—a technology that will revolutionize warfighting in ways that we cannot yet anticipate. But unlike nuclear technology, which spread slowly, AI is being rapidly and widely weaponized, with attendant dramatic changes in warfare.
What is to be done? The liberal-democratic West (including Japan, Korea and Taiwan) have an effective monopoly on the best AI hardware, especially for training giant models. The West should now cut off China, Russia, Iran and North Korea’s (i.e., the CRINKs) access to this hardware, and cripple their capabilities for building their own competing hardware based on Western semiconductor technology. The next wave of AI improvement is going to be revolutionary, and there’s no reason to enable our enemies to benefit.
No comments:
Post a Comment