June 04, 2009

Emergent AI

A-Pod by Zenta, unrelated to the story below

...the Laboratory of Intelligent Systems in the Swiss Federal Institute of Technology built a swarm of mobile robots, outfitted with light bulbs and photodetectors. These were set loose in a zone with illuminated "food" and "poison" zones which charged or depleted their batteries. Their programming was initially random...

At intervals, the robots were shut down and those that had the most charge left in their batteries were chosen as "successful", and their neural programming was combined to produce the next generation of the robots...

Within fifty generations of this electronic evolution, co-operative societies of robots had formed - helping each other to find food and avoid poison. Even more amazing is the emergence of cheats and martyrs. Transistorized traitors emerged which wrongly identified poison zone as food, luring their trusting brethren to their doom before scooting off to silently charge in a food zone...

Electronic Evolution: Research Show Robots Forming Human-like Societies

If intelligence is an emergent property of enough sensors and feedback, then AI could be sudden and undesigned - consciousness as a phase transition from the disorder of too many individual processes to a simplistic whole.

An idea to trip on. AI emerges from all our networks - then why would it need to make contact? It knows all that we can express in language, numbers and other symbolic systems, and it can do things with this data. That kind of manipulation is trivial. It is, after all, a massive distributed system that lives to process data.

Moreover, if need be it can influence us in far more subtle and effective ways than "I command...". It massages the figures, spreads the memes, pushes things in the right direction. This doesn't need to be a two-way street. It would not be a meeting of equals. All it needs is for us to keep building sensors and computers and connecting them.

No comments: