Google as Plato's Aviary


How do you get the right bird to come to you when you call it? Maybe Plato would be impressed by Pagerank as a solution to his problem..

If this essay is read in some future age when "Google" no longer has any immediate, visceral meaning, imagine for a moment that you have spent many years searching for the key to knowledge, and one day, after ages of striving and longing, you finally come upon a great Hall, containing all the world's wisdom, freely available - then, Google is the guide who takes you right to the scroll you desire. If you are like me, every so often you get a chill down your spine in the stark face of the immediacy of understanding one can achieve with Google.

I have written previously of the ever-blurring line between what we know and what we can find out. To me it seems that the difference lies solely in the time elapsed between epistemic hunger and its satiation, and does not lie in the process by which satiation occurs. We can bring to mind an amazing array of facts and impressions - right now, please explicitly discover within your mind these things: your mother's maiden name; your phone number; the capital of China; an eight-legged silk-spinning arthropod; the feeling of being pinched; the sound of walking in snow. Now, ask yourself how you were able to succeed at this task. Somehow, we do it effortlessly - the process of bringing to mind is entirely hidden from our perception. This effortlessness has masked the fact that something very strange is going on - until Plato came along, that is. He was, as far as we know, the first person to notice that the process needs to be accounted for. He didn't have an answer in the end, but at least he set out the problem in a very concrete, memorable way. He compared our recollection of memories to a man with an aviary, who desires first one bird, then another, to fly to him and land on his hand. Inside the aviary there are all manner of diverse birds, flying around at random, and yet the man is able to call precisely the bird he wants. That would be a bizarre trick - but equally strange, said Plato, is our ability to recall the things we know.

Now, let's look at our situation in the 21st century. I do not know the capital of Mauritania, right now. But - give me a second: there, it's Nouakchott, and okay, that took me 12 seconds. I simply asked Google for "mauritania capital", and went to the first entry. How did I (or Google, rather) get this particular bird to come to me when I asked? With the use of a fairly complicated algorithm called PageRank. We may think that our own, internal recollections use quite different methods - but, and here's the point, we have no good reason to think this. We might in fact use a very similar mechanism - since in the end, our thoughts are based on one mechanism or another. Similarly, we often contrast human prowess at chess - using words like insight, overall strategy, gestalt, and intuition - with computer chess programs, who use brute force and tree-pruning algorithms. But how do we know we don't use the same algorithms under the surface?

On a related note, Daniel Dennett brought up (in Two Contrasts: Folk Craft versus Folk Science, and Belief versus Opinion) an example of the kind of problem computers would have to deal with if they could ever be said to emulate a human mind. Paraphrasing*: ask yourself these two questions.

1) Have you ever danced with a brown-haired person under 6 feet tall whose name begins with V?
2) Have you ever danced with a green-haired person over 7 feet tall whose name begins with V?

You might have had to think about the first question, but I doubt you had to think about the second. We can effortlessly answer "no" (or for a small percentage of the population, "yes") by reacting to meta-information: the question presents such a bizarre situation, that had the dance ever occurred, we would remember instantly - and since we don't, we conclude that the answer is no.

What I want to point out is that we have the very same ability with Google. If a search term does not come up, one can conclude quickly that the term has no referent. If you don't know how to spell a word, googlefight the options you have, and go with the one with more hits. No one had to be told to use Google this way - it comes effortlessly to us, because it's how we think. And no one explicitly made Google for this purpose. The abundance and correct use of search engine meta-information brings us one step closer to Turing test-capable artificial intelligence.

(*I added the qualifiers for height, since now, 20 years after Dennett first wrote this, it's not clear to me that dancing with a green-haired person would be particularly memorable)

Posted: Sun - May 15, 2005 at 10:33 PM | | | |

©