The Preparedness Team
Under the leadership of MIT AI professor Aleksander Madry, OpenAI has established a dedicated team called “Preparedness.” Comprising AI researchers, computer scientists, national security experts, and policy professionals, this team will continuously monitor and test OpenAI’s technology. It plays a crucial role in identifying and alerting the company if any of its AI systems demonstrate potentially dangerous behavior. The Preparedness Team operates alongside OpenAI’s existing Safety Systems team, which focuses on mitigating issues like embedding biased algorithms into AI, and the Superalignment team, which investigates preventing AI from harming humanity in a hypothetical scenario where AI surpasses human intelligence.
The Existential Threats Debate
Earlier this year, prominent figures in the AI industry, including leaders from OpenAI, Google, and Microsoft, emphasized the existential risks associated with AI, likening them to pandemics or nuclear weapons. However, some researchers argue that such a concentration on hypothetical catastrophic scenarios diverts attention away from the current harmful consequences of AI technology. A growing number of AI business leaders advocate for moving forward with tech development, asserting that risks are overstated, and the technology can significantly benefit society as well as drive profitability.
A Balanced Approach
OpenAI’s CEO, Sam Altman, acknowledges the long-term risks inherent to AI but stresses the importance of addressing present challenges. Altman opposes regulations that disproportionately burden smaller companies and hinder their ability to compete in the rapidly evolving AI landscape.
Mitigating Risks with Expert Collaboration
Aleksander Madry, an esteemed AI researcher and director of MIT’s Center for Deployable Machine Learning, joined OpenAI this year. Despite recent changes in leadership, Madry expresses confidence in the company’s commitment to investigating AI risks. OpenAI is actively engaging in discussions with organizations like the National Nuclear Security Administration to facilitate comprehensive studies on AI risks. Furthermore, OpenAI pledges to permit external “qualified, independent third-parties” to assess its technology and conduct rigorous testing beyond standard research available online. In addressing viewpoints that polarize AI development as either accelerating or decelerating, Madry emphasizes the oversimplification of such debates, advocating for a more nuanced approach to AI advancement.