The Intelligence Age
Article

Interaction patterns for truth-seeking

published on · updated on

The previous article in the sequence is Reasons to worry about AI safety via debate.

Rather than focus on the particulars of the debate game posed by AI safety via debate1, here I will ask questions about how the debate game fits into a broader context of interaction patterns. With this in mind, I pose four questions:

  • Given a goal of seeking the truth in a multi-agent setting, what interaction patterns are available?

  • What are the common elements among these interaction patterns?

  • What are some ways to formalize the these interaction patterns?

  • How does one evaluate the quality of an interaction pattern?

The next article in the sequence is Downsides of debate.

Endnotes

1

G. Irving, P. Christiano, and D. Amodei, “AI safety via debate.” arXiv, Oct. 22, 2018. doi: 10.48550/arXiv.1805.00899.