What is the fastest way to reliably align a powerful AGI around the safe performance of some limited task that is potent enough to save the world from unaligned AGI?
What is the fastest way to reliably align a powerful AGI around the safe performance of some limited task that is potent enough to save the world from unaligned AGI?
Links
[1] https://www.edge.org/annual-question/what-is-the-last-question
[2] https://www.edge.org/inthenews/what-is-the-last-question
[3] https://www.edge.org/contributors/what-is-the-last-question
[4] https://www.edge.org/responses/what-is-the-last-question
[5] https://www.edge.org/memberbio/eliezer_s_yudkowsky