Just last week Udi Manber, Google's VP of Engineering gave a presentation "Search is a hard problem". The top three reasons cited by him were
- Scale and diversity are almost beyond comprehension
- Expectations and needs will continue to grow
- 20 to 25% of the queries we see today, we have never seen before
Considering that Google processes billions of queries and still 20 to 25% new queries is a huge number. Consider this to the route taken by Mahalo, "Human Powered Search Engine". The NYT article on Mahalo says
Mahalo now has pre-prepared pages for 5,000 terms related to entertainment, travel, health, technology and other subject areas. The company plans to expand its coverage to 10,000 terms by year-end, and eventually to provide results for one-third of the most common search terms.
The common search term is a factor of time and its is a moving target. Paris Hilton might be the top search term on Mahalo today,but that will change in a quick second with another breaking news.
- Can a Human Powered Search keep up with the changes in the real world ?
- Can it correct itself if a page is broken ?
- What about personalization ?
- What about languages and user location.
To be called a search, Mahalo should provide at least these bare minimum functions before it can try to improve on Google's results. Otherwise one would have to categorize Mahalo as a directory and not a search engine.
At Mahalo and other “human powered” sites, the machine performs a first cut at the search in advance of a user’s request, and the results are then winnowed and shaped by human editors, then stored, a process that Mr. Calacanis terms “editorialized search.”
This feels a lot like the the "Google Answers" experiment, where experts were supposed to help you with searches. Users are smart enough to switch search engines if there is a lot of webspam on one of the searches and compare a combination of results from Yahoo, MSN, Ask and Google.