Aug 14

Tim O'Reilly

Tim O'Reilly

Flickr and Interestingness

Recently, I was looking for a photo to illustrate a talk I was giving at a company meeting. I was telling the old story about three men working. A passerby asks them what they're doing. The first man grunts "working" and goes back to his stonework. The second man says, "I'm building a wall." The third man stops, gazes off into the distance for a moment, and says, "I'm building the most beautiful cathedral in Ireland."

I wanted to remind people what we're really doing each day. Not just working, not just building a wall, but focused on what Jim Collins calls a Big Hairy Audacious Goal. (O'Reilly's is changing the world by spreading the knowledge of innovators.)

So I needed some good pictures as visual bullets. I first went to Google image search, but while the photos were good, they were all somewhat "expected." After paging through the first four sets of results, still hadn't found anything suitable. So I went over to flickr, set my search to "most interesting" rather than most recent or most relevant, and soon had a fabulous selection of unusual photos of cathedrals.

Google made a breakthrough in web search with its original idea of links as citations (i.e. PageRank), and they are still the undisputed leader in general web search, but they haven't done as well in searching rich media. I think they have some things to learn from Flickr. More specifically, web search innovators all need to think through what makes results "interesting" for a given domain. I like what flickr has done in calling out "interestingness" as a quality worth searching for, and leaving it as a playground for exploration.

tags: lazyweb  | comments: 17   | Sphere It

Previous  |  Next

0 TrackBacks

TrackBack URL for this entry:

Comments: 17

  Till [08.14.06 05:52 AM]

I basically agree, however, the problem Google tries to solve is a little bit more challenging, since the photos are not provided (uploaded) by users and not tagged.

Guessing from web-site context on the content of embedded pictures is much harder.

Interestingness in flickr probably stems from number of comments, number of people who counted the picture in question as favorite and so on. All features that rely on specific user contributions.

We had similar bsevrations when we built a content-based image retrieval engine (

  Brian Sawyer [08.14.06 06:56 AM]

Flickr also makes it easy to find images to reuse:

Though this search filter would be a lot handier if the results could be sorted by interestingness.

  Kevin Farnham [08.14.06 08:46 AM]

I know and know of a lot of people who spend lots of time searching for interesting images for their MySpace profile pages. Until now I've advised them to use Google. Since the goal of putting images on your MySpace page is to make a page that is interesting to others who share similar interests or have similar artistic appreciation, Flickr would seem to be a much better place for performing this type of image search.

"Interestingness" is a useful concept, but can it be measured and quantified using an automated process? MySpace has a kind of interestingness gauge in the number of friends you have. Google in essence quantified interestingness by tying it to the number of incoming links, then enhancing the rating by assessing the importance of each incoming link through analysis of the global data set. In both cases, human action sets the initial condition (people create links on sites and add MySpace friends), which can then be "read" by a machine and put to some use.

Flickr sounds very interesting, much more than a simple image repository...

  Andrew Koper [08.14.06 02:54 PM]

I know this is OT to your main point, but my first reaction as I started to read your post - maybe thinking from an entrepreneur's perspective - was the following. From '97 - '02, I went to a lot of after work Internet events and conferences and had a lot of exciting conversations with a lot of smart people who all worked in the biz and wanted to make the Next Big Internet Thing. None of them became really sucessful, influential or famous.

Vision and "thinking outside the box" are good, but I always thought that the key to success was _develop_an_app_ (or _develop_a_useful_app_), market it, and sell it. Intelligent people who wanted to do something huge and thought about it and read a lot of blogs, Web sites, and magazines and went to conferences never did it - imagining is meaningless in ways.

-Andrew K
Detroit, MI

  Search Engines WEB ♣ [08.14.06 07:30 PM]

Flickr's selection will usually be more INTERESTING than Google's, simply because, many who upload to flickr, are uploading images they are quite proud of aesthically and want to share with others who would voice apprecation for the eyecandy.

Even sites like Corbis are searched for their WEB versions of their print images - and will treat users to some of the most alluring, captivating photos ever.

Many Website images on Google - were not always intended to be aesthetically pleasing by their creators. And some are quite compressed, because their purposes have to factor in fast loading pages.

The Urls, the ALT text and nearby body text will determine the Google IMAGE Algos rankings for photos - while tagging is quite significant in Flickr.

  Alex Andronov [08.14.06 11:53 PM]

I believe interestingness works by a combination of the following things:

1) How many people did something after seeing a picture eg. added a comment, marked it as a favorite. If there are two pictures which have the same number of favorites the one that has been less seen seems to be more interesting. Which although it seems counter intuitive is probably right. It shows which photos people have been moved by.

2) Favorites seem to have more weight than comments

3) I believe that each user has a score which is based on the number of interesting photos they have uploaded and that when they make a comment, add as a favorite or ignore a photo they do so with a weight which is given by this score. So a person who is known by the system to create interesting content is given greater power to judge content as interesting.

Obviously these are guesses based on the output rather than a knowledge of how it works.

With Google's Picassa online albums they might be about to wade into this territory but they have always been very weary of pissing in the gene pool.

A regular search example might be derived from this thought: What if every time somebody searches for "French Fondue Kits" they click on the eighth item on the list not the first. With a combination of Ajax and cookies they could record all of the followed links, and tell when you stopped searching. If everyone was doing it then that eighth link should slowly start moving up the page regardless of its "pure" page rank. But I don't believe Google does this.

Flickr pissing in the gene pool would be if they revealed your personal score. Then people would try and do things to affect it and then it would loose its value.

  Ian Delaney [08.15.06 09:49 AM]

Really thought-provoking topic. The Interestingness rating is formed by amalgamating the number of comments/views/references, is it not? So it is a photographic equivalent of Google PageRank or the Technorati Ranking. Does this really work for aesthetic judgements is the question. It seems to have done so for you. But I wonder. Is Mondrian more "interesting" than Picasso, or vice versa?

  bryan [08.15.06 02:41 PM]

one interesting thing about the three men working anecdote is that later it turned out that the man who grunted "working" ended up building the most beautiful cathedral in Ireland.

also I would like to say: argument by anecdote is a load of shit.

  Tim O'Reilly [08.15.06 08:56 PM]

bryan, interesting comment, given that you're arguing by anecdote. And your anecdote is indeed thought-provoking. We don't need to embrace a big vision to be caught up in fulfilling one, but it helps.

Take Google. When everyone else was loading up their search engine home page with advertising, Google steadfastly refused to do so, even though they had no other obvious source of revenue. Why? They had a vision. They pursued it, relentlessly focusing on what would be best for their users, and eventually discovered how to make money while fulfilling their vision.

Take Microsoft. Bill Gates had a vision of ubiquitous personal computing, which he pursued relentlessly. (He did a lot of other things relentlessly too, but that's another story.)

Take Apple. Steve Jobs has always had a vision of what he wanted to produce, which has led him to reinvent the company again and again.

  eve [08.21.06 11:11 AM]

computers that can see are getting more likely everyday. flickr is an interesting experiment, but it still relies on the willing and benevolent community for its "usefulness" (probably even weirder to define than "interestingness") as Alex pointed out. I was hoping to point you guys in this direction for another experiment that isn't quite so reliant on a community willing to leave helpful tags. peek and boom isn't striving to catologue things by interestingness but it's amazing nonetheless: i'm not sure if that (or this!) is ot or not, but i am fascinated by the synaptic-style networks enabled by tools such as flickr and myspace. another one i am quickly becoming unable to live without is stumbleupon. if the internet is a parallel universe, we should be able to map it infinitely.

  Rune Wiggers Ecklon [08.23.06 11:14 AM]

Google is doing things the robot way while the majority of web 2.0 services are going the human way. Google has yet to fully venture into services grown and build by humans but I doubt it will take long. Its natural for a company that has success in doing things a certain way to keep doing it that way until its a proven concept to do it in a different way.

Flickr is one of such proven concepts.

  Tim O'Reilly [08.23.06 12:05 PM]

Rune -- I don't agree. Google's original search insight with PageRank is one of the breakthrough applications of human intelligence to something that had formerly been thought of as purely an algorithmic problem. It's more subtle than many of the explicit applications of user-generated content, and as a result, perhaps more important. Looking for possibilities of implicit involvement of humans in computer processes is going to be more powerful even than explicit invitations to participate.

  Lars [08.29.06 08:53 PM]

That's a useful distinction: implicit vs. explicit human input. So part of the genius of Google PageRank and Flickr's interestingness is taking something that users would do anyway, and leveraging it as a source of valuable information beyond what each user knew they were doing.

  Thomas P. Jones [05.11.07 07:09 AM]

I think both Google's PageRank and Flickr's Interestingness Score drive the evolution of both the users and the systems.

As I strive to create images that score higher on Flickr's Explore pages, I advance my own knowledge, skills and social network. As my social network of contacts grows, I improve my ability to get pictures onto Flickr's Explore page and as a result get a higher score. Fame breeds more fame.

The same is true for the Flickr system. The ability of their Interestingness Score to pick out eye-popping photos drives my desire to see the pictures that make it onto the Explore page everyday. Now, I look to it to provide both inspirations AND admirations.

Also, I find the best way to GET more comments and favorites on my pictures is to GIVE more comments and favorites on the pictures other users share. It amazes me how much personal effort Flickr's Interestingness Score has driven me to and how rewarding it has been for me.

  Juande Santander [05.16.07 09:10 AM]

Flickr score seems to be available, and still content gets created —more or less— in the same way. You can experiment with Flickr inspector and a Flickr userid or a Yahoo! account:

And while Thomas P. Jones' are essentially on the mark, I'm starting to think we're putting too much effort on online communities, instead of human to human communities…

  Ryan [11.17.07 08:49 PM]

I hope you were only looking at creative commons images; crazy as it may sound, a good ol' powerpoint presentation does indeed require a license.

Post A Comment:

 (please be patient, comments may take awhile to post)

Type the characters you see in the picture above.