Posts Tagged ‘search engine’

Multiplex Division

13 May 2010

Dear J-

I’ve been thinking about Google lately; if you remember before the browser wars, there were competing search engines all around, from the first efforts of Jerry Yang’s Yahoo! to various spidering sites like Inktomi. The one that I used reliably roughly fifteen years ago was Altavista — this probably because I had a soft spot in my heart for Digital Equipment Corporation, at one point the number two vendor of computer equipment behind IBM and dedicated to going it on their own path — from the semi-PC compatible Rainbow 100 to the speed-is-everything Alpha chips — before being swallowed up by Compaq.

Google changed that; from the first inklings of how broad it was to the fascinating depths it could plumb, it soon became my default go-to engine. When I was little I learned that the term ‘google’ meant the number one followed by a hundred zeroes (1E100, if you rather), and it felt like the secrets of the Internet were laid bare in an unassuming little text-entry box. Although no one’s quite replicated the secret sauce that brings back bushels of hits, the current commercials from Bing are amusingly spoofy: there’s still a lot of work to be done to separate the chaff from the wheat, although a lot of people prefer doing just that.

I suppose that’s part of their success: everyone’s developed their own algorithm to sort through Google results; better to include everything rather than feel as if some robot somewhere is deciding what’s important. It’s funny that we’d accept that, but then again this is the land of super-sizing and SUVs: better to get more than you might need: you’ll grow into those shoes, kid. It works for search engines (as wide a net as you can cast is never a bad thing) but now Google wants to be Microsoft (everywhere, all the time) and I’m not convinced that their corporate motto (“Don’t be evil”) and generic bigger-is-better philosophy is sufficient any more.

Mike

Advertisement

Rule Based

9 October 2009

Dear J-

Some of the early efforts at artificial intelligence involved rule-based decision trees; the idea was that you’d end up adding a rule that covered every situation eventually and then the simulation was complete.  For instance, you could ask what the square root of 169 is; the computer would parse the request and figure out the answer is 13.  This is actually similar to how the Terminator and RoboCop were supposed to work — several scenes from the machine’s point of view had a number of appropriate choices scroll up in a sort of heads-up display.

I love these commercials for bing, Microsoft’s search engine, where “search overload” has made spouses into non-sequitir-spouting babblers; the fame of Google was built on its search engine, which used a proprietary algorithm to rank search results.  This was after early efforts at search engines on the web — let’s say AltaVista and Inktomi — proved quite efficient at sending out spiders to index page content; the next logical step was to prioritize those contents based on something better than exact phrase match.  We’re still stuck on figuring out the best way to rank pages — Google’s algorithm of scaling rank based on the number of pages linking to the one being ranked has been publicized enough that it’s relatively easy to manipulate.

It’s the downfall of any rule-based system; once you understand the rules, it’s simple to game.  When I was little my world ran pretty close to an AI decision tree:  in school, when I was faced with having to come up with a project for Science Fair or pick out decorations for our shoebox covered wagons I was completely stumped — and it showed.  I remember boiling leaves and saying something about chlorophyll (that was the teacher’s idea for a project, not mine), and in handing back the wagons, we were treated to an offhand lecture about how disappointed she was about some students taking home their wagons the night before and throwing sticks and cotton balls inside to simulate firewood and pillows (I peeked inside mine as she said it:  twigs, cotton balls, AND paperclip horseshoes).  The problem with rules is they don’t allow for nor do they nurture creativity; once I learned that rule I realized how restrictive the rest of them were.  I’m still picking my way through the resulting wreckage.

Mike