Cuil Your Jets: Take Offs Are Easier than Landings
July 28, 2008
Digging through the rose petals is tough going. Cuil, it seems, has charmed those interested in Web search.
Balanced Comments Are Here
Among the more balanced commentaries are:
- Search maven Danny Sullivan here who says, “Can any start-up search engine “be the next Google?” Many have wondered this, and today’s launch of Cuil (pronounced “cool’) may provide the best test case since Google itself overtook more established search engines.”
- Michael Arrington, TechCrunch here, says, “Cuil does a good job of guessing what we’ll want next and presents that in the top right widget. That means Cuil saves time for more research based queries.”
- David Utter, WebProNews here, says, “The real test for Cuil when it comes back will be how well it handles the niche queries people make all the time, expecting a solid result from very few words.”
Now, I don’t want to pull harder on this cool search stallion’s bit. I do want to offer several observations:
First, the size of indexes don’t matter. If I am looking for the antidote to save a child’s life, the system need only return one result–the name of the antidote. The “size matters” problem surfaced decades ago when ABI / INFORM, a for fee database with a typical annual index growth of about 50,000 new records, found itself challenged as “too small” by a company called Management Contents. Predicasts jumped on the bandwagon. The number of entries in the index does not correlate to satisfying a user’s query. The size of the index provides very useful data which can be used to enhance a search result, but size in and of itself does not translate to “good results”. For example, on Cuil, run the query beyond search. You will see this Web log’s logo mapped to another site. This means nothing to me, but it shows that one must look beyond the excitement of a new system to explore.
Second, the key to consumer search engines is dealing with the average user who types 2.3 terms per query. The test query spears on Cuil returns the expected britney spears hits. Enter the term britny, and you get very similar results, but the graphics rotate plucking an image from one site and mashing it into the “hit”. Enter the query “brittany” and you get zero hits for Ms. Spears, super star. The fuzzy spelling logic and the synonym expansion is not yet tailored for the average user who can spell Ms. Spears more than 400 ways if I recall a comment made by Googler Jeff Dean several years ago.
Third, I turned on safe search and ran my “brittany” query. Here’s what I saw in the inset that allows me to search by category.
I like Playboy bunnies, and we have dozens of them hanging around the computer lab here in Harrods Creek. However, in some of the libraries in the Commonwealth of Kentucky, a safe search function that returns a hutch of Playboy bunnies can create some excitement.
Fourth, it is not clear to me what learnings from WebFountain, Dr. Patterson’s Google patent documents, and Mr. Monier’s learnings from the AltaVista.com/eBay/Google experiences have or have not found their way into this service. Search is a pretty difficult challenge as Microsoft’s struggles attest over the last 12 or 13 years. My hunch is that there are some facets to the intellectual property within Cuil that warrant a lawyer with a magnifying glass.
I applaud the Cuil team for getting a service up and running. Powerset was slow out of the starting blocks and wrangled a pay day with a modest demo. Cuil, in a somewhat snappier, way launched a full service. Over the coming weeks and months, the issues of precision, recall, relevance, synonym expansion, and filters that surprise will be resolved.
I don’t want to suggest this is a Google killer for several reasons. First, I learned from a respected computer scientist that a Gmail address set up for a test and not released seemed to have been snagged in a Cuil crawl. Subsequent tests showed the offending email address was no longer in the index. My thought was that the distance between Cuil and Google might not be so great. Most of the Cuil team are Xooglers, and some share the Stanford computer science old school spirit. Therefore, I want to see exactly how close or far Cuil and Google are.
Second, the issue of using images from one site to illustrate a false drop on another site must be resolved. I don’t care, but some may. Here’s an example of this error for the query beyond search.
If this happens to a person more litigious than I, Cuil will be spending some of its remaining $33 million in venture funds to battle an aggrieved media giant. Google has learned how testy Viacom is over snippets of Beavis and Butt-head. Cuil may enjoy that experience as well.
To close, exercise Cuil. I will continue to monitor the service. I plan to reread Dr. Patterson’s Google patent documents this week as well. If you want to know what she invented when working for the GOOG, you can find a eight or nine page discussion of the inventions in Google Version 2.0. A general “drill down” notion is touched upon in these documents in my opinion.
And, keep in mind, the premise of The Google Legacy is that Google will be with us for a long time. Cuil is just one examples of the Google “legacy”; that is, Xooglers who build on Google’s approach to cloud based computing services.
Stephen Arnold, July 28, 2008