mediajunk

In their quest for Artificial Intelligence (AI), many computer scientists first aim for a child-like AI with the ability to learn from reading and experience.

As Stanford University Professor of Computer Science Dr. John McCarty puts it:

This idea has been proposed many times, starting in the 1940s. Eventually, it will be made to work. However, AI programs haven't yet reached the level of being able to learn much of what a child learns from physical experience. Nor do present programs understand language well enough to learn much by reading.

But what if computer "minds" could learn a different way? Not by reading but by Googling?

In his lecture All I Really Need to Know I Learned from Google, Oren Etzioni (University of Washington) describes how his research group's AI project uses Google to discover facts about the world.

Called the KnowItAll project, the machine "learns" by searching Google (and, more recently, its own web index) for certain types of phrases, then analysing the results.

For example, let's say KnowItAll wants to find a list of town names in Ireland. It could do this is by searching for the following phrase:
"Irish towns such as"
- and examine the words that follow this phrase, in each of the web pages in which it occurs.

It could also search for "and other Irish towns", and identify the words that come before this phrase.

Repeat this formula with many other useful phrases and you get a list of words that are likely candidates for towns in Ireland. Words that occur with less frequency, or do not occur in all of the lists, are less likely to be town names.

I'm glossing over the details here. The methods and algorithms used for finding and determining facts is much more complicated but, in principle, this is how it works.

What does this mean for the future? Hmmm... Well, within milliseconds, your artificially intelligent friend could find out the answers to each of those tricky Pub Quiz questions. And you wouldn't even have to buy it a beer. Thanks, Google School!