Elementry My Dear Watson

In true dork fashion, last week instead of watching the  match-up between “Watson” the supercomputer and Jeopardy’ s top two players, I watched a Nova episode on it. After getting to the underpinnings of just how it’s software works, a few thoughts came to mind. Bear with me.

• The quality of data that Watson uses is hugely important. If it’s premises are false then so will be it’s answers.
• Apparently, IBM programed the computer to learn patterns. But, what kind of patterns does it look for? Patterns aren’t as simple as we would like to think, because they’re bound by the rules of what they are – conceptual, physical, or otherwise.
• In stances where Watson can’t decide, “he” relies on probability. The software can’t override this function.

So, what does this mean?

Essentially, it appears that IBM has created a super search engine that integrates voice-to-text technology with the ability to “learn.” It’s not intelligent, but it certainly pushes the boundaries of what that means. The question in my mind is, “what’s next?

Intuitively, the next step is problem solving. Yes, Watson relies on textual cues to solve them already, but he can’t answer them without this context. Moreover, he can’t create information from scratch.

While search and retrieval are important elements when answering questions, being able to answer questions often requires creating new information. Or, a taking non-textual data into account.

Case in point – Hubble and the expanding universe. Prior to Hubble’s use of redshift to determine that certain objects in the sky are far away, astronomers thought that the Milky Was was the only galaxy in existence.  But, Hubble’s ability to 1) collect raw, primary data led him to b) create unique information that used redshift to prove that distances between celestial bodies are often vast.

Watson is limited to using text. He has no ability to call into question the status quo, and without this, no new information can be produced.

I’ve gotta admit, though, I like the idea of a real-time speech driven, computational search engine. And, if the answers are given in terms of probability then it’ll let me do my job even better than now.