- WATCH - PARTICIPATE on
If 2001: A Space Odyssey's iconic HAL 9000 set the bar for you as to what defines a thinking machine, well, just a decade late, that bit of science fiction has become reality. The TV game show Jeopardy features a computerized contestant this week, an array of servers designed by IBM and answering to the name "Watson."
First, the creepy: It wasn't enough to challenge Jeopardy's top alumni to a man-versus-machine faceoff, so the engineers, or set designers, or someone, decided to stand-up a plasma screen at the podium between the human contestants. The monitor displays animated graphics representing the 'puter which pulse along with its artificial voice ...bringing us to the icing on the cake: Instead of a GPS-sounding stutterfest, Watson vocalizes its responses with a smooth cadence that's just a little too sedated for my peace of mind.
The theatrics are outstanding, and truly worthy of a little time on your DVR, but the movie-like user experience of Watson is only a part of what is happening here. Unlike its Hollywood counterpart, there is no actor reading lines behind Watson's facade. The machine is actually accepting Alex Trebek's vocal prompts, interpreting them, searching its data banks (it's not currently connected to the Internet), physically activating a buzzer mechanism, and rendering an audible response when called upon (in the form of a question, per the rules of the game).
Remember, on Jeopardy, clues are rarely simple or direct. Just deciphering what the host is saying can sometimes take mere mortals a few seconds, in and of itself. When prompted, "A classic by Crockett Johnson, Harold and the blank crayon," Watson won the buzzer and correctly answered, "Purple" What is intriguing about that exchange is the sophisticated parsing of the english language into binary code that Watson was instantly able to achieve.
The science behind turning human inquisitiveness into actionable computer instructions is the secret sauce of your favorite web search engines. When you "Google" something, software algorithms first try to determine what you are looking for and then return a ranked list of the most closely related information. So when Watson hears "Harold and the blank crayon," determines that "blank" really means, "What is the missing word in this phrase?" finds Crockett Johnson's book title, identifies the missing word "Purple" and returns that as its answer, well, I think that's pretty darn intelligent.
Just for fun, search the web for George Bush Senior's infamous utterance using the term "Read my blank" and see how well the best-funded algorithms on the planet do. A while back, AskJeeves.com, set out to create a common-speech web search, hoping people would eventually be able to type in search queries the way they talk and get meaningful results. Essentially, the company programmed its engine to ignore unnecessary words like "why" or "the" and search as normal. It didn't yield any meaningful advancement, and everyone has since copied the practice. So, when Watson stands apart from the pack, we should recognize it for the accomplishment it is.
It seems a little anticlimactic, though, doesn't it?
By simply playing a game, this machine teaches us that artificial intelligence doesn't require some exotic new architecture or nuclear power plant to support its faster-than-light processor. It just needs to be able to talk to us. In the end, we didn't need an omniscient monolith of cybernetics, we just needed a better translator in the digital world.
Instead of a computer that "knows" everything, it will probably be our hands-free dialing that effectively changes the world. When you can speak plainly to your technology, the interface with humanity becomes less and less apparent, and when we connect tomorrow's super-search engines to the world wide web, they will seemingly always know what they are talking about, and probably do as well as any human does in casual conversation.
The next step will be software that can intuit what you are thinking. Using cues from body language or inflection, a suite of apps in the future could offer a range of automated reactions designed to aid and comfort humanity, all without any real intelligence at the helm. Will that be enough? Probably not.
Researchers will likely continue
to pursue the creation of a sentient artificial intelligence. Creating
self-awareness is the Holy Grail of AI science, and whether or not it's
actually possible is probably a question that ventures into the domain
of philosophy more so than computer science. Either way though, Watson
is much more than I believed was possible in today's mundane world,
and the possibilities it opens up are very exciting, if a little unnerving