Saturday, December 01, 2007


My pick of the week's top 5 items of interest:
  1. The academic community has shot themselves in the foot once again: Two researchers from UT Austin have written a paper showing how to de-anonymize the anonymous 100 million+ movie ratings Netflix made available to the public in 2006. While this is an impressive feat, it will likely make companies that have access to rich data sets much less likely to release them to researchers in the future. While this may be a positive thing for privacy advocates, it's certainly not for researchers. What these UT guys have done in effect is demonstrate why academia should not be trusted with data they desperately need to do their research.

    Update: A friend of mine emailed me about this issue and made some good points: Maybe Netflix should have offered some prize money to figure out how to properly release anonymized data. Also, researchers have an obligation to show how others might misuse "anonymous" data... companies making such data available to the public should show much more care in the future.

  2. We're all carrying ticking time bombs in our pockets... UPDATE: Or maybe not.

  3. More bad news for Microsoft: a recent benchmark test revealed XP was twice as fast as Vista. And in court, MS's marketing director was just as confused about what "Vista capable" means as the rest of us.

  4. Philipp Lenssen has written an interesting article about Abstract vs Realistic Icons, making some really good points about how making icons more photo-realistic doesn't necessarily improve usability.

  5. Colin N., a former CS student, emailed me about a new games contest sponsored by Microsoft for students. I'd like to change the was the CS software development class works by targeting this competition next year.