🎉 MIT FutureTech has a new paper out, and I’m grateful to have been part of it. This Delphi study brought together 272 experts to ask which AI risks are most severe, who is most vulnerable, and who is most responsible.

A few findings worth sitting with:

🟣 A responsibility gap: the people most vulnerable to AI risk (e.g., everyday users and the public) are the least responsible for fixing it. Responsibility sits upstream with developers and governments, while the most exposed have the least power to act. 🟣 Under business as usual, 18 of 24 AI risks are rated as carrying a ≥10% chance of catastrophic outcomes (1M+ deaths or $100B+ in losses) by 2030. 🟣 Most severe expected harms: dangerous capabilities, competitive dynamics, weapons & cyberattacks, power centralization, and false information.

🔵 Finally, as someone working at the intersection of AI fairness and animal welfare science, I really appreciate the paper’s acknowledgment that AI systems’ harms to non-human entities like animals weren’t directly included, and are likely undercounted by many experts.

This research is part of the MIT AI Risk Initiative. This work was led by Alexander Saeri, Jess Graham, and Michael Noetel 🔸, Peter Slattery, PhD, and Neil Thompson. Special thanks to Peter Slattery, PhD for inviting me!

🔗 Interactive findings: https://airisk.mit.edu/priorities 🔗 Full paper: https://cdn.prod.website-files.com/669550d38372f33552d2516e/6a172558bd2947234379749f_a8684052fd49a64374c9a9d3e4e5ab59_Prioritizing%20the%20risks%20from%20Artificial%20Intelligence.pdf

#AI #AIrisk #ResponsibleAI #AISafety #AnimalWelfare