Many times over the last few years I've made attempts at lectures giving an overview about what kinds of futures for humanity the unfolding saga of AI development may lead to, and about what is being done within the still small but rapidly growing field of AI alignment with the goal of improving the odds for a benign outcome. Some of these attempts are available in video format in earlier Häggström hävdar blog posts and elsewhere. They have varied in quality, and in a few instances I've walked away from the podium with something resembling a sense of having nailed the topic as best as I can hope to, but to tell the truth I have never ever come even close to the level exhibited by the brilliant and inimitable Scott Aaronson in the talk hosted by AT Austin's EA (Effective Altruism) group he gave a couple of weeks ago, on November 14, 2022. Warmly recommended!
The most vivid and pedagogically useful explanation of the near-term dangers implied by the development of AI technology for military drones that I know of has been the famous Slaughterbots video, produced in 2017 by the Future of Life Institute and featuring leading AI safety researcher Stuart Russell. I've been showing it to my students every year since 2018, along with proclamations such as ''Potential dangers in AI lurk everywhere, but if I were to single out one application area as the absolutely most reckless one as a playground for development of cutting-edge AI technology, I'd say killer robots''.
Just this week, two event have come to my attention that cause me to update on this and perhaps modify my lectures slightly:
1. A large part of the Slaughterbots video is a fictional promotion event for a new autonomous killer drone product. But why stick to fiction when we can have the real thing? The Israeli defense electronics company Elbit Systems has released a promotion video for their latest lethal AI product. Their tone and content are reminiscent the Slaughterbots video but without any hint of irony or admission that AI technology for killing people could be dangerous or bad. Here's what we get when reality imitates fiction:
2. Facebook's (or Meta as they now call themselves) AI lab just announced how they built an AI that performs on (at least) human level at the game of Diplomacy. Given the role of Bostrom's treacherous turn concept and related deception problems (as outlined in the recent very important report by Ajeya Cotra) in exacerbating AI existential risk, I should perhaps consider modifying the proclamation above to something like ''Potential dangers in AI lurk everywhere, but if I were to single out one application area as the absolutely most reckless one as a playground for development of cutting-edge AI technology, I'd say games where the winning moves consist of manipulating human opponents into acting in your interest and then stabbing their back''.