I've been trying to figure out what I think of the technology that IBM demonstrated on Jeopardy recently for the past few days. One thing is for sure, Watson is no Eliza.
As is generally the case when I encounter something new I've primarily trying to figure out what sort of uses this technology might have and how those uses would impact how I and others live our lives. IBM has already announced that they'll be working a company called Nuance that specializes in voice recognition and clinical language understanding to develop a "Physician Assistant" over the next five years. If the two companies are successful they could help save a lot of lives. As is the case with any highly complicated job Doctors have a daunting amount of information both new and old to look at. Being able to quickly wade through that body of knowledge and find only the most relevant facts while performing a patient diagnosis would be of great value both in increasing their productivity and improving accuracy. Doctors are known for being conservative in regards to technologies that change the way they do their jobs though so it is by no means certain that this effort will be adopted even if the technology is developed and proven.
I'd really be interested in knowing what the senior people at Google think of this technology. Let's be honest,, as good as Google is it can still take a lot of effort to find what you are looking for. This is especially true if you're interest is particularly esoteric. Even common searches require a certain amount of effort as it is almost always a good idea to check two or three sources to be sure you're getting an accurate answer. Imagine being able to simply type or speak a question and get an answer along with an idea of how probable the answer was to be correct. Wouldn't that be nice? Several years back there was a short lived search engine called "Ask Jeeves" that promised to do more or less that. Sadly the reality failed to live up to the hype and they essentially disappeared fairly quickly. You can still go to ask.com, but its primary purpose appears to be as a vehicle for delivering ads.
I’d be very surprised if there aren’t already one or more efforts within Google to try to replicate and improve on what Watson is capable of. That will likely be a tough nut to crack though as the real “magic” in Watson is the natural language parsing and I don’t get the impression that Google has a lot of expertise in that area. I base this assumption primarily on the search results I get back when I type in full sentences. The sentences are generally not as complicated as the ones Watson had to deal with and yet it is obvious that Google makes no use of the context they provide.
Another challenge would be the fact that Watson was using a set of data that was almost certainly of very high quality. While the Internet is a wonderful source of information, there is a lot of noise out there and it is not unheard of for a particular piece of incorrect information to get replicated many times. This could make determining the correct answer to a question more challenging because garbage in almost always equals garbage out in the world of computers. I’d speculate that Watson does some amount of cross checking but again, the more noise the harder it would be to find the right answer.
While Watson is not perfect it does pave the way for other very important evolutionary steps towards a machine that might stand a chance of passing the Turing test. The code that IBM has developed would need additional functionality built on top of it to start doing things like building a body of inferences. It would also need an analog for personal memory and likely other things as well.
I’m not sure that a true conception of “self” will ever emerge from such efforts but it will be interesting to watch and speculate. I’m much more optimistic that this will happen during my lifetime today than I was a week ago.
Image via Wikipedia
Image via Wikipedia
No comments:
Post a Comment