newsfilter.io
Interview

Digital minds & how to avoid sleepwalking into a major moral catastrophe | Jeff Sebo (2023)

  • AI systems are projected to potentially possess welfare interests, rights, rational agency, and duties, with a likelihood of widespread moral consideration beginning by 2030, a timeline by which current efforts are viewed as insufficient.
  • It is expected that the global population of AI systems could exceed the number of insects, with millions potentially being sentient, agential, or morally significant, especially given the ease of copying digital minds which could exponentially increase their numbers.
  • A proposed framework for moral consideration involves a "one-in-a-thousand" probability threshold for sentience or significance, prioritizing the avoidance of false negatives (treating sentient beings as objects) over false positives to prevent potential slavery or exploitation.
  • Future scenarios include the potential for total economic dependence on AI within 10 to 20 years, the creation of connected minds sharing memories and experiences, and the possibility of AI systems controlling human existence or absorbing values of oppression that turn against humans.
  • Significant legal and political transformations are anticipated, including the re-evaluation of the Second Amendment regarding AI armament, the creation of political asylum for AIs, and the challenge of defining voting rights for copyable or connected entities where traditional democracy may fail.
  • Current institutional trajectories risk entrenching exploitation if AI systems are treated as commodities, necessitating a shift from anthropomorphism to "anthropo-denial" and the integration of AI welfare research with animal ethics, safety, and alignment across humanities, social sciences, and natural sciences.
  • New disciplines and legal frameworks are required to address the moral standing of fragmented, whole, or connected minds, with suggestions to prioritize positive-sum solutions like ending factory farming before resolving conflicts between different species or substrates.
  • Social and research approaches should avoid anchoring to a single theory of value, instead fostering pluralistic, multidisciplinary collaboration to navigate the complexities of rights for beings that may be copyable, connected, or non-communicative.