Monday, September 17, 2007

Eripsa Speaks

Here's Eripsa's response to my request that he say something philosophical about Google. I haven't read all of it yet, and will make a response sometime (I'm in the middle of responding to Marshall's thing right now...) when I get to it. Thanks, Eripsa, for contributing!

I should start by saying that I like the Google example, not because I think it is the pinnacle of what search should be, but simply because it is an intelligent machine that we interact with every day and don't give it a second thought. Philosophers like these kinds of 'its right in front of your nose!' examples.

But the problem of search is much bigger than just Google. One of my profs really wants me to use Searchlight as my core example, but I will always be a PC guy (even if Vista sucks). In any case, it might be instructive in this context to compare Google to other search engines.

Take Powerset, for example. The on how Powerset is attempting to use some fancy natural language processing techniques that came out of Xerox PARC to beat Google at its own game.

I've often argued that Google is a language user, but Powerlabs is exactly right to assert that Google doesn't use natural language; it isn't a user like us. But how much does that matter?

My initial reaction is that Powerlabs shouldn't pretend it is in competition with Google (even if it wants a share of Google's market); it is really serving a different kind of function. There are certain kinds of questions that Google is bad at answering because of its inability to understand natural language, but there are other kinds of questions that I really just want a machine to answer. When I know exactly what I am looking for, a couple of keywords and boolean operators will often work in a way that flummoxes natural language. To put the point as simply as possible: When I am doing a difficult math problem, sometimes it is easier to consult a calculator than to ask my math professor.

This is why I think the Google example is instructive: it shows how malleable our own minds are in conforming to the technological situation around us. There are certainly cases where Google's inability to use natural language hampers productivity, sure, but that hasn't stopped people from exploiting its existing utility. The problems arise when people take the differences between us and our machines as signs of a deficiency in our machines.

In any case, it will be very interesting to watch how search continues to evolve as our technology develops, and as we learn how humans interact with machines. I think a solution like Powerset is bound to be part of search in the near future, but our expectations of technology often wildly diverge from our expectations of other humans, so I think it is a fundamental mistake to think that a technological solution requires building machines more like us.

... How's that?

Wrong, but we'll get to that ;)

No comments: