A hold on AI: on the Preliminary Report of the Independent International Scientific Panel on AI.
AI holds scientific promise, but it should not unfold unchecked.
The UN’s Preliminary Report of the Independent International Scientific Panel on AI drives at a few fault lines in the rapid investment into and proliferation of AI technologies: the Global South-Global North divide, with the latter poised to take the lion’s share of the benefits of the diffusion of advanced AI models in different industries; and the challenge poorer countries face in regulating models far more advanced than what their own AI ecosystems can develop. These divides force emerging countries to choose between capital-intensive undertakings to get a seat at the table, or to accept the hand dealt to them by the half-dozen companies whose decisions shape AI access and use. These divisions are all the more urgent when seen in the light of what makes a country a voice that matters in AI: abundant electricity, highly capitalised firms that employ scarce talent with high salaries, and a drive to plow through policy resistance with a competitive zeal to outdo last month’s capabilities. While the report highlights AI technologies’ transformative potential in scientific research, its development is far more rapid than that of social media — an industry that had time to act responsibly yet deformed public discourse and affected democracies, even as people were more or less able to tell what was going on in real time. It is unclear whether individual countries, even China or the U.S., have enough power to meaningfully arrest how AI develops.
AI firms acting irresponsibly have already caused harms that would not be tolerated in any other industry. AI models have ensnared teenagers and adults in parasocial fantasies that have sometimes turned fatal; they have flattened the world wide web, grievously injuring the news media’s ability to deploy resources in their respective missions to inform the public; they have unleashed an epidemic of deepfakes, undermining trust in the written word and images alike; and they have deeply intertwined their promises of Artificial General Intelligence with global financial systems, with possible catastrophic economic consequences. The most important task for governments is how to hold these firms to account when needed, as taking a backseat on this conversation — even if industry leaders ignore the Global South’s concerns — carries far greater costs. India has already experienced the cost of not asserting itself in the technology development of the decade: even as Anthropic’s Mythos and Fable held out the tantalising promise of securing vast cyber systems, the U.S. pulled access, leaving the firms that had these models to ponder the risk of a different model succeeding at attacking their infrastructure. There is a limit to visions of AI as a geostrategic asset; much needs to be done to address its potential for broad damage.