AI Safety Wargame Exercise (Optional Prep Session)

However it develops, Artificial Intelligence (AI) is without a doubt a highly influential phenomena. Due to the scope of its potential consequences and the neglect of AI safety relative to that scope, the safe and beneficial development of AI is currently considered an important cause area by many in EA. (See for example, https://​​80000hours.org/​​problem-profiles/​​positively-shaping-artificial-intelligence/​​)

For a fun and challenging exploration of the possible consequences of AI, we’ll be conducting a wargaming exercise called Intelligence Rising (https://​​intelligencerising.org/​​).

This is a scenario role-play game developed and run by researchers from Cambridge, Oxford, and Wichita State University. It is designed to expose players to high-stakes decision junctures faced by powerful actors (states and corporations) as AI systems become more general and their societal impacts grow, and it offers a space for participants to test their assumptions and understanding of AI strategy and impacts.

The game has participants take on the role of powerful stakeholders such as tech company CEOs and leaders of countries with the potential to affect the future of AI. A game lasts approximately four hours, and play takes place in turns, with participants asked to announce their public actions, plan secret actions, and enter into negotiations with other participants. Each turn takes the participants some years into the future. The participants must also contend with external factors, determined in part with the help of a random number generator.

The game is suitable for 4-16 participants (ideally 8-12), and we can run multiple sessions for a larger audience. Player experiences are optimal if they participate in an optional one hour preparatory session one week before the game.

Welcomingness:
This event is open to all new or returning members of EA Toronto community, and no background knowledge is required. Participants with an existing interest in the development of artificial intelligence will get the most out of the event.

We encourage interested participants to skim the topic archives for AI Policy on 80,000hours.org: https://​​80000hours.org/​​topic/​​priority-paths/​​ai-policy/​​

No comments.