Re: Military Friendly AI

From: Brian Atkins (
Date: Sat Jun 29 2002 - 16:48:06 MDT

Eugen Leitl wrote:
> Is anyone here interested to co-author a short paper on risks in
> Singularity AI, and suggested regulations?

I'm not at the moment since I believe this should be an issue of self-
regulation, but one idea:

Any project announcing that it is working on a "seed AI" type design
with even tiny chance of self-improving takeoff should be willing and
able to publicly post and defend a "social policy". This policy should
at a minimum describe the following:

1. The project's experiment protocol, from the design of the AI to
containment, etc. The works.

2. How this team plans to tackle the "long term Friendliness issue"

3. Anything else that I forgot

This should all be available no later than 6 months before running full
code so the public at least has a chance to counter with any perceived
errors in the plan.

Brian Atkins
Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT