Ask HN: IBM will kill Google with Watson?
Now that IBM has a computer that can answer complex human language questions, beating the world's best Jeopardy contestants, what's stopping IBM from feeding Watson the entire Internet as his knowledge base, and giving him a front end website for user's to ask him questions?
Forget searching pages of Google search results. Watson could answer nearly anything you ask it in real natural language.
Sure it'll take a year or two more to refine Watson to scale, but all the hard work is done. Additional using his machine learning, Watson would get better and better as user's asked him questions. What say you Hacker News? Will IBM kill Google?
14 comments
[ 2.8 ms ] story [ 36.2 ms ] threadAt the risk of repeating myself, trivia questions are a game, not a source of wisdom. They are a highly constrained genre: a good trivia question requires no special expertise, can be answered in seconds by a human, and - most important of all - has a single, unambiguous, indisputably correct, short answer. In other words, it is the least important form of knowledge. They call it trivia for a reason.
It is doubtful that Watson will be particularly better at answering such questions than Google plus a half-educsted human would be. It might be faster than that combo, but how often are fractions of a second of the essence when Googling? It might not require the human to do as much, but humans are a dime a dozen on Mechanical Turk.
Moreover, taking humans out of the search loop is self-defeating: Google makes money because a human must filter the results. Humans are susceptible to advertising. Watson won't click on ads that he glimpses out of the corner of his eye, and even if he did advertisers wouldn't pay for those clicks. So what is WatsonGoogle's business model?
If half seconds weren't important in search, Google would not have developed instant search.
What I dispute, however, is the point of the original suggestion: that Watson represents some kind of massive conceptual leap over Google, one that could translate into a threat to Google's business.
Today, Watson has two advantages over Google's search engine: It plays Jeopardy -- which is a very impressive demo; it's hard to be impressed by Google search anymore now that kids have grown up with it -- and it gets to work on one question at a time using a profoundly expensive pile of special-purpose hardware. These things are awesome, but they are not actually magical. By the time Watson's tech becomes available on the web for everyone, Google may well have it too. By the time I can afford my own pocket Watson, Google may well be selling a competitor.
Also, I could be wrong, but I imagine a large percentage of searches are for trivia. It wouldn't have to be all or nothing. Perhaps Watson answers trivia like questions in clear english. For everything else he returns all the resources he has that he thinks are related to the question, like Google does. Heck, maybe Google partners with IBM to license the tech. Google's already trying to answer some trivial questions outright in search, such as when you ask for a definition, unit conversion, or simple math. Watson could take that to a whole new level.
As far as a business model, I don't see why slapping an ad on where to purchase the correct size sockets you need online beside your answer wouldn't work.
Using Google or Watson is an inherently inefficient approach because it's trivial to find the most practical answer with a socket set and a mechanics creeper (and '77 Corvette). A shop manual doesn't hurt either.
IBM has little to no interest in or tolerance for consumer facing web sites. It tried developing a consumer web search in 1996 called "InfoMarket" and killed it weeks before launch (and killed off IBM's own web site search in the process as collateral damage).
While IBM does support a few sporting event web sites (which IBM has down to a routine), I don't see IBM doing anything with Watson facing the general public. Perhaps as a gimmick for a short period of time, but not a general service.