Last week produced two starkly opposed visions of artificial intelligence and its potential overlap with geopolitical negotiations. In San Francisco, one of the world’s leading-edge frontier AI labs called for the sector to slow down its innovation. And in Buenos Aires, Argentina’s president proposed having AI robots run companies without human oversight or management. While neither story will likely play out to the full extent of their respective announcements, the contrast demonstrates a geopolitical challenge for the coming decade: Efforts to regulate artificial intelligence will face spoilers who will constantly push the boundaries of any regulation that is implemented.
The call to slow down came from Anthropic, the AI company behind the Claude chatbot. In a recent report, it warned that AI systems are advancing so rapidly that they are approaching a moment of “recursive self-improvement,” in which the systems can design, build, train, and even deploy their own upgrades, with humans only marginally in the loop.
There are plenty of reasons to exercise caution as countries and companies compete to produce and deploy higher-quality AI models. Those concerns include potential AI-enabled cyberattacks against critical infrastructure systems, the economic shocks that would result from AI displacing millions of workers and heightened geopolitical competition as the U.S. and China race to obtain the best models that may provide a military advantage. They also extend to low-probability but catastrophic events that have until now been the province of science fiction, such as superintelligent machines trying to destroy humanity. As usual, political debates are slow to catch up to emerging technologies.
Leave a comment