The AI Dream

Had an odd dream this morning. Out of the blue, haven't been thinking about this topic, so it wasn't the result of anything I recently discussed, or read or saw in a movie of late.

I was in a small group watching a research demonstration of a new type of AI that was developed for some security application. A robot, in the form of a toddler, was turned loose to search for something on a stage in front of us. I was impressed how quickly he was moving around obstacles and under a table searching and commented to a gal next to me that if that was a real toddler he would have already whacked his head on that table at least once.

Then the toddler got a little confused and wandered off the stage and started interacting in some interesting ways with others and his "extended" environment. Ways that didn't appear to be the result of any programming that I could imagine.

So I asked the researcher if he had noted signs of emergent behavior like this before in his experiments. Which he vehemently denied. He argued that this was a controlled experiment and didn't have the complexity of the natural world where emergence (while still controversial) applied. I argued that any subset of the natural world was still part of the natural world and that the "noise" in the data of any "controlled" experiment indicated it wasn't possible to hermetically seal this relationship. We agreed to disagree, and I confessed that I wasn't completely objective about this as I had already made up my mind that emergence existed.

Weird. Emergence is one of those curious properties in things that arise and can't be explained or predicted by taking those things apart. Sort of flies in the face of our reductionist version of reality currently in vogue among scientist who haven't grokked the new paradigms. But it's time for my morning coffee. Maybe then it will make sense.