Wednesday, August 13, 2008

When good software goes bad

I basically like Google as a company. Yes, they've grown huge, but I think they've learned from Microsoft's mistakes and aren't squashing competitors like bugs or locking in customers with draconian licensing terms. For the most part, Google seems to be more good than bad. Actually, much more.

One of the little free services they offer is a customized news page--- you may have played with it: http://news.google.com/ You can set up a page of results with your specific topics of interest covered. Google uses your search history to tailor the search results towards the kinds of news stories and sites you've previously gone to.

It's updated every few minutes, so the news is very fresh and up-to-date. It's almost like having a headline scan of a good and continually-updated newspaper at your fingertips. (And of course, a click or two brings you out to the actual, full coverage of any given story.)

The Google software that assembles your page even tries to find little illustrations to accompany major stories. The illustrations sometimes are directly from a linked story, but if the story has no such available graphic, Google will find a suitable graphic from among its database of image links.

It usually works quite well. But sometimes, there are unintended consequences:

Yesterday, for example, some medical news revealed that large numbers of younger adult men are getting unnecessary prostate cancer screenings. The Google software looked at the keywords in the news article and then did a search through Google's vast repository of image links, looking for a photo with the same keywords in the same relevance and frequency.

Presto! Image found! News item assembled and published!




(click to enlarge)


OK, maybe the software still needs a little tuning.

4 comments:

  1. It's funny when that happens, yes. :D It's funny when it sometimes picks a picture of the wrong celebrity for a story sometimes.

    BTW, your URL for news.google.com is invalid - you've used \\ instead of // after the "http:" bit.

    ReplyDelete
  2. The http://news.google.com link that I see points to http://www.blogger.com/%5C%5Cnews.google.com. The slashes are ok it's the anchor that's wrong.

    ReplyDelete
  3. That's so funny. I saw that image too and thought "I guess you have to train somehow"

    ReplyDelete