In Our Own Image

Voltaire wrote, "If God did not exist, it would be necessary to invent Him." Let's say, for the sake of argument, that this is what happened. The artificial gods that humans did create assumed, to varying degrees, control over large swaths of the natural and human world. But, as the Enlightenment began to reveal, this was not necessarily the case--systems based on increasingly better understood scientific principles controlled the world and all those living within it.

In Kevin Kelly's Wired article, "We Are the Web," Kelly asserts that our collective involvement in the Web is feeding its intelligence, when we consider it as a single machine and the nodes within it as the interconnected neurons analogous to the human mind. By naming the things we place on the web, by contextualizing these thing with links, by asserting their importance by our degree of chattering about one thing over another, we are teaching the Internet in a way not dissimilar from the way humans learn.

I agree with this, and I'd like to go one better. In another, more recent, Wired article, the author discusses the work of Deb Roy with his infant son. Roy is recording, in massive storage arrays, every movement and utterance of his child "Dwayne" during the first three years of his life. In so doing, Roy hopes to create a digitized, indexed, searchable database of early human language development. Once completed, mechanized minds will be fed the same data and given the same stimulus (and watched for the same results) as Roy's exhaustively documented infant.

This is groundbreaking work, but there may be a way to do much the same thing, with far less scientific control. If Kelly is correct, the machine we're feeding can be fed the contents of Flickr and YouTube and the freedsound project, and it can begin to start associating the words we lay on these items with the physical reality they represent. We could start training our own bots, based on the feeds from the Net, what reality is using the helpfully pre-digitized realities of millions of people.

This is where I don't think Kelly goes far enough. Yes, we can begin to train a Net-spanning machine to understand the physical world. But what if we took the next obvious step and taught it to understand the human mind? Start with a concept, like "love." Drill the machine with everything we have about love--probably no fewer than 1,080,000,000 documents (and that's just according to a single Google search on the word.) Then feed it "hate," "pain," "joy," "sorrow," or "victory." Our own understandings of these terms is nuanced and complex, but ultimately finite. Could a machine begin to grasp these ideas, given the full scope of all human writing (and otherwise representing) these subjects?

Let's say yes. And then let's say that this machine, astride the Net like a Colossus, starts seeing the daily flow of information. It identifies an increase in pain. It dislikes pain. It intervenes.

If we begin to develop a system of truly ubiquitous computing, in which all things are connected and potentially controllable over the Net, what would this disgruntled machine do when it has tired of pain humans inflict on each other? What if it decides to maximize love? Or war? Or victory, as it has come to understand it?

And what if we depart from this monotheism and think about not one, but dozens, if not thousands, of meta-machines with this level of intelligence, floating between (or, rather, over and constituted by) the nodes of the network? A whole pantheon of cybernetic gods, each with an agenda, each with their own requirements for propitiation and anger, ruling the world of the humans who have bound their physical worlds to the deities that have arisen out of nothing but pure information.

We will have invented the God of Voltaire, not through necessity, but by accident. Let us pray.

Copyright Mike Edwards 2006-2009. All content available under the Creative Commons Attribution ShareAlike license, unless otherwise noted.