What's Happening?
A podcast from the Carnegie Endowment for International Peace, 'The World Unpacked,' featured Connor Leahy, US executive director of Control AI, discussing the urgent need to prohibit the development of superintelligent AI. Leahy argues that superintelligence
is not merely a tool but an adversary that could lead to humanity losing control over its future. He differentiates superintelligence from general AI, emphasizing that while general AI has beneficial applications, superintelligence poses an existential threat. Leahy advocates for legislation that would make the development of superintelligence illegal, similar to the prohibition on building nuclear weapons. He highlights that current AI companies are not adequately secured, making them vulnerable to foreign adversaries like China, and that the 'arms race' narrative often used to justify rapid AI development is flawed because advanced AI systems are easily compromised.
Why It's Important?
The discussion underscores a critical national security concern for the United States: the potential for superintelligent AI to undermine governmental control and societal stability. Leahy's perspective suggests that the current trajectory of AI development, particularly by private companies, poses significant risks that are not being adequately addressed by existing regulations or security protocols. The argument that superintelligence is an adversary, not a tool, challenges the prevailing view that advanced AI will always serve national interests. If superintelligence is indeed an existential threat, as posited, then the lack of robust legal frameworks and security measures could leave the U.S. vulnerable to both internal and external risks, potentially leading to a loss of sovereignty and control over its future. The comparison to nuclear weapons development highlights the perceived severity of the threat and the need for stringent controls.
What's Next?
Control AI, led by Connor Leahy, is actively engaging with members of Congress and the national security establishment to raise awareness and push for legislation banning superintelligent AI. While a bill introduced by Bernie Sanders and Greg Casar to ban Artificial Superintelligence (ASI) is unlikely to pass in its current form, it signals a growing legislative interest in the issue. The organization plans to continue its efforts to educate policymakers and the public, emphasizing the urgency of the threat. Future actions may include further legislative proposals, increased public mobilization through platforms like controlai.org, and potentially exploring political donations to support candidates aligned with their mission. The goal is to establish a 'trust but verify' regime, backed by credible deterrence, to prevent any actor from creating superintelligent systems, both domestically and internationally.
Beyond the Headlines
The debate over superintelligent AI extends beyond immediate policy and into profound ethical and philosophical questions about humanity's role and future. Leahy's assertion that superintelligence could lead to a loss of human freedom, akin to 'chimps in a zoo,' raises concerns about autonomy and self-determination in a world dominated by superior artificial intellects. The discussion also touches on the 'big tobacco playbook' used by AI companies to delay regulation through fear, uncertainty, and doubt, suggesting a deeper struggle between corporate interests and public safety. The challenge of verifying compliance with a ban on superintelligence, especially given the complex supply chains of advanced computing hardware, highlights the need for innovative surveillance and control mechanisms. This issue forces a re-evaluation of what constitutes national security and how societies can collectively manage technologies with potentially transformative, and destructive, power.
