The Lifecycle of Software ObjectsOver the weekend I read Ted…

The Lifecycle of Software Objects

Over the weekend I read Ted Chiang‘s novella The Lifecycle of Software Objects. (links to get this book down in the last paragraph… it’s complicated*). 

This novella is the most compelling vision of how near-term truly intelligent AI will evolve that I’ve ever read. Ted paints a world in which there are no short cuts to building a truly intelligent AI, no special math tricks or heuristics that can be accelerated by faster hardware.  Instead, in order to build AI successfully, one must raise a genetic algorithm from infancy, and put in the necessary time via social interaction to teach the code proper values and morals, very similar to raising a child.

Assuming that AI progress will follow Moore’s Law just because the underlying hardware follows Moore’s Law is naive, in Ted’s view.  This one paragraph encompasses the lessons from the book well, and are Ted’s words taken from the epilogue:

People routinely attempt to describe the human brain’s capabilities in terms of instructions per second, and then use that as a guidepost for predicting when computers when be as smart as people.  I think this makes about as much sense as judging the brain by the amount of heat it generates.  Imagine if someone were to say, “when we have a computer that runs as hot as a human brain, we will have a computer as smart as a human brain.” We’d laugh at such a claim, but people make similar claims about processing speed and for some reason they get taken seriously.

It’s been over a decade since we built a computer that could defeat the best human chess players, yet we’re nowhere near building a robot that can walk into your kitchen and cook you some scrambled eggs.  It turns out that, unlike chess, navigating the real world is not a problem that can be solved by simply using faster processors and more memory.  There’s more and more evidence that if we want an AI to have common sense, it will have to develop it in the same ways that children do: by imitating others, by trying different things and seeing what works, and most of all by accruing experience.  (emphasis added)

If you’re enjoying the macro conversation in the tech landscape today about the progress of AI, I strongly encourage you to read Ted’s novella.

* Unfortunately this book is out of print and not available via Kindle for some reason, but you can find the full text on the Subterranean Press website. It’s missing a short section of author commentary at the end of the book, which is actually really valuable. This epilogue is published in blog post over here. To get the epilogue and the book together, you’ll have to grab a torrent copy of the epub.