Welcome! These forums will be deactivated by the end of this year. The conversation continues in a new morph over on Discord! Please join us there for a more active conversation and the occasional opportunity to ask developers questions directly! Go to the PS+ Discord Server.

Keeping AGIs confined

3 posts / 0 new

Last post

Tue, 2011-10-18 12:21

Arenamontanus Arenamontanus's picture

Keeping AGIs confined

Video where my colleague Stuart explains some of the issues involved in keeping super-intelligent seed AGIs confined: http://www.youtube.com/watch?feature=player_embedded&v=Gz9zYQsT-QQ The paper mentioned can be found here: http://www.aleph.se/papers/oracleAI.pdf Of course, from a role-playing perspective most of these methods and problems are instead possible adventure inspirations!

—

Thu, 2011-10-20 06:09

King Shere King Shere's picture

Re: Keeping AGIs confined

One quick idea is that some "minor" disaster at a forgotten sewage plant caused great mayhem. Namely that certain mesh components, various AGI & AI self terminate due to inbuilt safety protocols (a secret deadman switch at the site screamed bloody murder). The Fall should also have triggered such "unfortunate" responses. "Unfortunate , the droid suicides are not due to a hacking, software error or a mechanical malfunction, its a safety protocol response." Remembering the loose starting plot of a a old spy film where a seemingly unimportant department was cut, but the closing of that department caused brainwashed sleeper agents to awaken and start performing acts of terror. It was to their "all is safe, false alarm, go back to sleep" condition .

—

[img]http://boxall.no-ip.org/img/tc_userbar.jpg[/img] [img]http://boxall.no-ip.org/img/exh_userbar.jpg[/img] [img]http://boxall.no-ip.org/img/argo_userbar.jpg[/img] [i]"To find fault is easy; to do better may be difficult."[/i] Plutarch

Thu, 2011-10-20 07:05

Arenamontanus Arenamontanus's picture

Re: Keeping AGIs confined

Heh, overly fragile failsafes is a great topic! After the Fall most big AGIs were likely designed with a failsafe that crashes them if they show signs or even intentions for self-enhancement, including various secret trapdoors. Which of course suggests a few ways of doing denial-of-service - just inject "suspicious" information in their dataspaces, triggering the failsafe. Similarly the "excessive" safety protocols are great things for players or NPCs to get annoyed by. Why should the infomorph character have that blackbox that records their thoughts in case they ever misbehave? And the corporate system is so cumbersome to use, no wonder people greyboxed it.

—