Twenty-three Asilomar principles for Artificial Intelligence and the Future of Life

Abstract

The “Asilomar, California Conference on Beneficial AI” was a conference organized by the Institute for the Future of Life (Future of Life Institute) in January 2017, where over a hundred experts and researchers came together to discuss and formulate principles for ethical AI. The twenty-three principles are divided into themes or research questions (five); issues concerning ethics and values (five) and long-term problems (five). This chapter addresses each of the principles, makes them explicit and, finally, analyzes and discusses their feasibility and validity. It is concluded that the movement has received both acclaim and criticism. While some have hailed the principles as a valuable reference point for discussions of AI ethics, others have expressed concern that they lack specific guidance on how to apply them in practice. The question arises as to whether these principles are sufficient to address the complex ethical challenges that arise with artificial intelligence. The answer is that while they might not be enough on their own, they are necessary to initiate and shape this debate.

Time pressure and instinctive responses to driving dilemmas

“…If a collision is unavoidable, then how a vehicle crashes and what it crashes into could be lifesaving (Lin, 2016), and an AV's actions becomes an ethical issue with an outcome determined by its algorithm (Goodall, 2014). AV algorithms are still undergoing development, and the limited consensus on which ethical standards AV algorithms should prefer presents a challenge in developing AVs with ethical standards aligning with the Asilomar AI principles' suggestions that machine ethics should align with human values (Morandín-Ahuerma, 2024). Utilitarianism corresponds to choices that minimize harm and maximize the overall good (Lin, 2016).…”