First Published on the Googlizarion of Everything
I spent most of yesterday playing around with Google’s newest competitor, .
I did a series of searches on Cuil and compared them with Google. The results, when I could get past the server crashes cause by the rush of others trying it out on its debut day, were very different. Some were better for me on Cuil. Others were better for me on Google. But the experiences (and results) were so different that it’s clear to me that the very familiarity and comfort with the Google experience has a major effect on users — even me. This is beyond “network effect.”
I have to admit that I am thrilled that Cuil has come along now. Its design and features help me explain much of what I have been writing about here and in the book. I hope it survives and establishes some market share, if for no other reason to demonstrate that there is more than one way to search the messy collection of documents that we call the Web.
Cuil does three things differently from Google:
1) Cuil offers a very different interface and layout of search results. Instead of the top-to-bottom ranking of results that often bear no relation to each other, Cuil offers two or three columns of entries that fit much better on the horizontal screens on which we all read. There is no clear hierarchy on each page of results — no numbers or anything. Cuil excerpts seemingly random chunks of text from each result page. The snippets play a bigger part in the user experience than they do in Google. Most significantly, Cuil places a box of related subject tabs on the right side of the page. This can guide users to common idea areas or specific items they might have failed to recall when doing the initial search. One of my biggest problems with Google Web Search is that it presents the illusion of quality via rankings. That linearity undermines the important yet frustrating aspects of research: you don’t always know what you are looking for or know what is best for you.
2) Cuil bases its searches on semantic relations and patterns, not user history or incoming links. This is a major break from the “tyranny of the majority” that is the central operational feature of PageRank. Certainly, Google has been working on folding in semantic measures of quality and relevance into PageRank for some time. But the folks at Cuil seem to believe in its potential as a replacement for the PageRank principle. This is refreshing. It means searches produce very different results on both engines. For instance, here is a common Google search for “Siva.”:
Note the prominence of yours truly. This is one of the absurd things about Google and PageRank. There are almost a billion Hindus in the world. Don’t you think that a search for the name of a Hindu god should be dominated by references to that God? Sure, I am one important Hindu! Don’t get me wrong. But I am not as important as any god, Hindu or not. Oh, searches for “Thor” on both services yield links to motocross and Marvel Comics, as there are very few Thor worshipers left in the world. It’s troubling that the once-relevant band Smashing Pumpkins and I get high billing in this Google search. That’s because Google’s PageRank privileges the cyberactive and motivated. So groups that tend to produce links — like Smashing Pumpkins fans and Sivacracy friends, tend to have their results show up high, regardless of relative relevance in the real world.
Here is the same search on Cuil:
As you can see, this is a somewhat better (i.e. relevant for most of the world) search. It has no references to me or my blogs. But instead of the chief Smashing Pumpkins fan site (called Siva), it has a guitar chord and tab site about the song “Siva.” However, Cuil fails to offer Wikipedia sites as high hits in many searches, including this one. This is a shame, as Wikipedia is a very helpful place to start reading about almost any subject (although it also has a cyberbias). Google, for more than a year, has placed Wikipedia pages very high in most search results. That’s a very good thing. Overall, comparative searches are sometimes better on Cuil, sometimes better on Google, but very different almost always.
3) As a result of the linguistic turn, Cuil has no need to keep a dossier on us to improve search. This is one of Cuil’s major selling points. And it certainly should make Cuil the search engine of choice for holligans, misfits, pottymouths, and n’erdowells of all stripes. But seriously, Cuil might be able to demonstrate that quality search need not depend on massive data collection and exploitation. If it fails, however, Google’s triumph will once again generate calls for stronger global privacy regulation. So watch this contest closely.
These three factors raise all sorts of interesting questions.
• How does Cuil “learn?” One of the few benefits of massive personal data collection is that Google can train its searches over time to conform to our clicking choices (even if they are initially limited by PageRank in the first place).
• How will Cuil demonstrate its alleged superiority in search? This will be tough, as there is no clear or universal standard of quality.
• How can Cuil deal with other languages and scripts?
• How can Cuil fold in images, video, maps, etc. to their search results?
• Why would an advertiser go with Cuil over Google if it does not offer targeted ads based on massive data collection and mining?
• Will Google do as it always does — welcome competition and swallow the best of its innovations?
• How can I raise my Cuil rank so I can remain “Internet famous” on every major search engine?