[ENGLISH ONLY] On August 31, 2021, Przemysław Pałka gave a guest lecture at Yale University (on Zoom), introducing the European Commission’s proposal of the Artificial Intelligence Act. He offered a critique of its most dangerous provisions and assumptions and proposed an alternative way to govern AI systems.

On August 31, 2021, Przemysław Pałka gave a guest lecture at Yale University (on Zoom), introducing the European Commission’s proposal of the Artificial Intelligence Act. He offered a critique of its most dangerous provisions and assumptions and proposed an alternative way to governing AI systems. Here’s the abstract:

In April 2021 the European Commission rolled out its proposal for the Artificial Intelligence Act; a first in the world attempt to comprehensively regulate all AI systems. The Act, if passed, will apply extraterritorially, and is expected to serve as a model for other jurisdictions. At the first glance, the AIA looks like a dream-come-true for scholars who’ve been calling for the AI’s regulation in the past couple of years. The AIA contains obligations for keeping a human in the loop, ensuring data quality, designing risk-management plans; it even establishes the regulatory agency for algorithms! However, as the talk will argue, the AIA’s impact will be far from positive. As a neoliberal and technocratic law at its core, the AIA masks political decisions as technical ones, and effectively delegates these decisions to private consultancy- and law firms. It will generate tons of documentation that no one will ever read, without meaningfully changing the practice of the AI systems’ design and use. And given its many exceptions and loopholes, the AIA will not only fail to reach the AI systems used by big tech companies like Facebook, Google, or Amazon but also preempt the member states from doing so in the state legislation. The AIA is not only a bad law; it’s a dangerous one.