Robot’s Rules Of Order v3

As Artificial Intelligence and Robots become part of our life, we need to define a set of universal rules of engagement. i

We need to have a crisp set of rules that will be eventually embedded as essential behaviors — a set of Robot’s Rules Of Order, very similar to Robert’s Rules of order ; a larger superset of Asimov’s Laws of Robotics.
In fact, in the last six months or so, I have had two iterations in this direction and in this blog am revising once more !
RRO v1 was a lot simpler
RRO v2 builds on v1 — thematically not that different
RRO v1 & RRO v2 were influenced by Google’s paper “Concrete Problems in AI Safety” and an article @slate on MSFT CEO Satya Nadella’s views “The Partnership of the Future

Recently, two more references shaped my thoughts and inspired this next revision - The Asilomar AI principles developed conjunction with the 2017 Asilomar conference (videos here), through the process described here and …

… an interesting paper about Open Ended Intelligence, the winner of the Kurzweil Prize for Best AGI Idea, at the Artificial General Intelligence Conference 2016.

All the four references are very insightful - you should read them when you are the sharpest, ponder about their relevance and debate their implications.
  • The Asilomar AI principles reinforce the rules of Algorithmic Accountability, Intelligibility (Failure and Judicial Transparency), Value Alignment, Non-subversion, Intelligent Privacy & the self-awareness of one’s idiocy.
  • Usually we think of AGI as “The ability to achieve complex goals in complex environments”. The OEI paper points out that “intelligence never starts with solving a problem but much earlier in the formation of the problems” — The art of becoming intelligent ! I like this diagram a lot.
  • OEI also encompasses the interaction of a collection of agents interacting together, modifying their behaviors — This is core to AI, especially Robots.
Agents overcome their initial incompatibility and become coordinated by constraining their own or each others’ set of possible behaviors as well as their connectivity.
In other words, they reciprocally determine or select each others’ behaviors. Mutual modification of behavior requires direct or indirect reflexivity among agents (feedback)

Coordinated-collective-reflexive behavior is the core to our civilization and should be an integral and structural part of the AI world as well - a behavior that is embedded in all Robots we create …

What says thee ?