Comments on: Interestingness vs. Meaning http://www.amyallcock.com/blog/2006/01/07/interestingness-vs-meaning/ Wed, 08 Feb 2012 02:28:53 +0000 http://wordpress.org/?v=2.7.1 hourly 1 By: Amy http://www.amyallcock.com/blog/2006/01/07/interestingness-vs-meaning/comment-page-1/#comment-24 Amy Mon, 09 Jan 2006 14:58:58 +0000 http://www.amyallcock.com/blog/2006/01/07/interestingness-vs-meaning/#comment-24 I agree with you that a computer can determine what people find interesting based on their viewing habits and that is exactly what the interestingness algorithm does. According to the Flickr site: <blockquote>There are lots of things that make a photo 'interesting' (or not) in the Flickr. Where the clickthroughs are coming from; who comments on it and when; who marks it as a favorite; its tags and many more things which are constantly changing. Interestingness changes over time, as more and more fantastic photos and stories are added to Flickr.</blockquote> As for meaning, I don't think that interestingness implies meaning in the artistic sense. There are lots of cute pictures of cats and sunsets on the interestingness page but I don't think that they have meaning as works of art. Although I think that an algorithm can make recommendations based on previously specified viewing patterns I don't think that an algorithm can determine aesthetic value or artistic metric due to the subjective nature of both of those qualities. Flickr might be able to extract from my list of favorites that I have a preference for abstract and minimal photographs but I don't think that it can determine what I would consider to be a good one or not. I agree with you that a computer can determine what people find interesting based on their viewing habits and that is exactly what the interestingness algorithm does. According to the Flickr site:

There are lots of things that make a photo ‘interesting’ (or not) in the Flickr. Where the clickthroughs are coming from; who comments on it and when; who marks it as a favorite; its tags and many more things which are constantly changing. Interestingness changes over time, as more and more fantastic photos and stories are added to Flickr.

As for meaning, I don’t think that interestingness implies meaning in the artistic sense. There are lots of cute pictures of cats and sunsets on the interestingness page but I don’t think that they have meaning as works of art.

Although I think that an algorithm can make recommendations based on previously specified viewing patterns I don’t think that an algorithm can determine aesthetic value or artistic metric due to the subjective nature of both of those qualities. Flickr might be able to extract from my list of favorites that I have a preference for abstract and minimal photographs but I don’t think that it can determine what I would consider to be a good one or not.

]]> By: Dan Zen http://www.amyallcock.com/blog/2006/01/07/interestingness-vs-meaning/comment-page-1/#comment-23 Dan Zen Sat, 07 Jan 2006 20:30:38 +0000 http://www.amyallcock.com/blog/2006/01/07/interestingness-vs-meaning/#comment-23 I think that there is a probability that on average, a computer can figure out what people will find interesting based on a past history of humans telling the computer what they found interesting. "Interesting" and "meaning" are perhaps two different things but the odds are that if it is interesting to someone, it has meaning. I mean, I think the odds are favoured in that direction. Marketing, etc. is certainly based on principles that our reactions are predictable in general. There are exceptions. But perhaps as exceptions are encountered, they will be factored in leading to less exceptions. Most people cringe at the thought of a computer telling you what you like and that has led to a slow uptake on personal agents. But they are becoming accepted like stores and music sites recommending related items. Amazon, Del.icio.us, etc. as you are well aware. I think FlickR is using shapes and colors too in their formula and I assume the parameters are based on what others have favourited. For example, if on average, people have recorded that they like pictures that have the bottom half almost a uniform color and that color is a light color then FlickR will continue to assume that any picture that matches this is interesting. It certainly can be wrong but it is kind of cool to find out why. And if done correctly, it should work most of the time or certainly beat the ods. If you only have so much time, and you are not going to see every picture, and the pictures are still varied, then I don't mind seeing a filtered set. Remember that computers are not the only ones who filter. Dan I think that there is a probability that on average, a computer can figure out what people will find interesting based on a past history of humans telling the computer what they found interesting.

“Interesting” and “meaning” are perhaps two different things but the odds are that if it is interesting to someone, it has meaning. I mean, I think the odds are favoured in that direction. Marketing, etc. is certainly based on principles that our reactions are predictable in general. There are exceptions. But perhaps as exceptions are encountered, they will be factored in leading to less exceptions.

Most people cringe at the thought of a computer telling you what you like and that has led to a slow uptake on personal agents. But they are becoming accepted like stores and music sites recommending related items. Amazon, Del.icio.us, etc. as you are well aware.

I think FlickR is using shapes and colors too in their formula and I assume the parameters are based on what others have favourited. For example, if on average, people have recorded that they like pictures that have the bottom half almost a uniform color and that color is a light color then FlickR will continue to assume that any picture that matches this is interesting. It certainly can be wrong but it is kind of cool to find out why. And if done correctly, it should work most of the time or certainly beat the ods.

If you only have so much time, and you are not going to see every picture, and the pictures are still varied, then I don’t mind seeing a filtered set. Remember that computers are not the only ones who filter.

Dan

]]>