Thursday, February 17, 2011

“I, for one, welcome our new computer overlords”

Well, I'm going to steal a quick news article. We may not have our warp drives or teleporters yet, but we recently got a good deal closer to the world pictured by Star Trek. In a special exhibition show on "Jeopardy!", an I.B.M. supercomputer named "Watson" defeated Ken Jennings and Brad Rutter -- two well-known Jeopardy champions. After realizing that Watson's lead was too great to overtake, under his correct Final Jeopardy answer Ken Jennings wrote "I, for one, welcome our new computer overlords" (taking the line from a Simpsons episode).

It's interesting living in in a world where Deep Blue can defeat Kasparov, and Watson can best Jennings; where we find directions from Google rather than from a map, and where someone can print and bind a book as cheaply as a library can lend it. (FYI, as I understand, it costs the Harvard library about $3 every time someone checks out a book (mostly for paperwork and re-stocking), but the Internet Archive bookmobile can download, print and bind a book for just about $1.)

I.B.M. is looking at using this question answering technology as a programmable physician's assistant. It's admittedly next to impossible to predict upcoming innovations like this, but I'd like to ask where else you see things like this leading? Where do you think this kind of technology will go, and how will it affect daily life? Is this technology going to go down as a major breakthrough, or do you see it as essentially Google / Wolfram Alpha with voice recognition? What else do you see computers doing in the near future?


  1. Bill,

    I am an Jeopardy fan. So, two observations: Humans do not press button until there is an inkling of answer, however, Watson can and use the time alloted to answer to seek answer in the database. Or, Watson can play a very risky game, most humans would not.

    Second, this is much bigger than Physician's assistant stuff IBM is peddling. We basically can now encyclopedic most all fields, not just words, but for patterns too... I think we are entering computational intution realm.

  2. Hey A1,

    Thanks for the response. If you look at the NY Times article, it's actually interesting how the "buzzer factor" worked with Watson. It wasn't like this super-fast Commander Data-style speed buzzing. It had a weighted scheme where it would buzz in more quickly if it was more confident in its answer. If it was less confident, it would buzz in somewhat more slowly. But even so, it wasn't instantaneous, and the human players were able to beat it to the buzzer from time to time. No one seemed to think it was an unfair advantage.

    I agree that this is going to be much bigger than the doctor's office. (That was the main application that I could find that IBM specifically mentioned.) It will be interesting to see where it is applied.

  3. I'm sure search engine technology could benefit.

  4. Doh, you stole my post!! I was about to post it in the next day or two. Perhaps, I'll post it anyway since I try to explore some of the AI issues involved.

    Good post though. I also welcome our binary slurping overlords!

  5. Bill, it was very impressive to watch. I would bet that the scientists that designed that computer will be getting job offers from Google, Yahoo and Bing to apply what they learned to search engines.

  6. It could be good to have it as an extension of the brain (Google is a bit like that), a bit like exoskeletons for the rest of the body; also it could be part of one.

  7. Bill,

    Thanks for NYT's report on buzzer. Yet,somehow, I am uncomfortable because a computer can play the risk game without side effects. A human player must pay chemical load price (racing heart beats, sweaty palms, etc.) for each and every buzzer push; Watson only demands for more juice (electricity) and gets charged up, again!

    I presume IBM guys were honest, but as speed and algorithms improve, would you fault Watson-X (tenth gen) from betting? Intimidating? I recall there was a very smart player who will go for highest dollar clue for each category. He did well, until his luck ran out!

  8. jmb275 -- Sorry I stole your post. I felt kind of bad because of all the people on this blog, I'm probably the least computer savvy.

    Joe -- I'd bet they'll be getting offers from all sorts of places, but I'd also bet that IBM will try to make it very worth their while to stay right where they are. Who knows. I'm sure this technology will disseminate out to lots of different sectors.

    Cartesian -- I can see a lot of people thinking in that direction, but even if the science gets completely worked out (which would be tricky to say the least), from a public policy standpoint I honestly doubt it would happen any time soon. Historically any sort of invasive human engineering projects have immediately met with so many ethical, moral, biological, and many other concerns (many rightly so) that they never really seem to go anywhere and get deadlocked in politics. Who knows. We'll see if anything like that ends up happening.

    A1 -- I imagine that people like Ken Jennings and other heavily-experienced competitors have worked at this long enough to minimize interference from their emotions, but it's true that no matter how well they work or how long they practice, they still have to deal with emotions and other basic biological factors at some level. A computer never will. I've heard that any sort of competition like this is mostly a mind game. It's hard to tell how those types of factors would come into play with a computer. I would be very interested to see what they come up with by the time of Watson-X.

  9. Bill,

    Jennings is an exception. Most players are good, but would be in the dustbins with Watson-I.

    Watson-X probably will control the masses, reminds me of Borgs of Startrek.

    The only saving grace is that humans learn much faster, and almost always self-learn, and fast.

  10. A1 -- I agree that Jennings is an exception. In fact, "Most players are good, but would be in the dustbins" with Jennings.

    I certainly hope this doesn't go the way of the Borg. I guess we'll just see.

    Humans have a lot going for them. Creativity, the ability to go beyond their original programming -- these are very human traits. Nonetheless, it's interesting to see what they can do with computers.


To add a link to text:
<a href="URL">Text</a>