philip lelyveld The world of entertainment technology

The Right Way to Regulate AI

... The United States does not need so many new AI policies. It needs a new kind of policymaking. ...

The problem with reaching for a twentieth-century analogy is that AI simply does not resemble a twentieth-century innovation. ...

Instead of reaching to twentieth-century regulatory frameworks for guidance, policymakers must start with a different first step: asking themselves why they wish to govern AI at all. Drawing back from the task of governing AI is not an option. ...

Technologies that went undergoverned are now hastening democratic decline, intensifying insecurity, and eroding people’s trust in institutions worldwide.

But when tackling AI governance, it is crucial for leaders to consider not only what specific threats they fear from AI but what type of society they want to build. ...

The Biden administration has begun to make moves to apply the same approach to AI. In October 2022, the White House released its Blueprint for an AI Bill of Rights, which was distilled from engagement with representatives of various sectors of American society, including industry, academia, and civil society. The blueprint advanced five propositions: AI systems should be safe and effective. The public should know that their data will remain private. The public should not be subjected to the use of biased algorithms. Consumers should receive notice when an AI system is in use and have the opportunity to consent to using it. And citizens should be able to loop in a human being when AI is used to make a consequential decision about their lives. The document identified specific practices to encode public benefits into policy instruments, including the auditing, assessment, “red teaming,” and monitoring of AI systems on an ongoing basis. ...

If policymakers return to first principles such as those invoked in the AI Bill of Rights when governing AI, they may also recognize that many AI applications are already subject to existing regulatory oversight. ...

It is much less common in the world of policymaking. But NIST’s use of policy versioning will permit an agile approach to the development of standards for AI. NIST also accompanied its framework with a “playbook,” a practical guide to the document that will be updated every six months with new resources and case studies. ...

Democratic leaders must understand that disrupting and outpacing the regulatory process is part of the tech industry’s business model. Anchoring their policymaking process on fundamental democratic principles would give lawmakers and regulators a consistent benchmark against which to consider the impact of AI systems and focus attention on societal benefits, not just the hype cycle of a new product. ...

AI governance need not be a drag on innovation. Ask bankers if unregulated lending by a competitor is good for them. ...

See the full story here: https://www.foreignaffairs.com/united-states/right-way-regulate-ai-alondra-nelson?utm_source=substack&utm_medium=email