I love the magnet like slopes - it's one that just makes you go "wow". Reminds me somewhat of the James Dyson "Wrong Garden" illusion as discussed previously on HN (http://news.ycombinator.com/item?id=779555, comments of thread discuss how to see a short video of it).
The "counter-intuitive illusory contours" one is interesting. It is almost as though our brain is treating the image as a liquid, and imagining distortion waves ahead of the "collision point".
The sine wave one is interesting in that you no longer notice the effect if you are reading the text below it (i.e not focusing on the wave beyond your periphery). This is somewhat similar to the "spinning dancer" illusion (http://en.wikipedia.org/wiki/The_Spinning_Dancer) in which you can trick your mind to see the illusion in different ways, and sometimes have difficulty seeing it again as you did originally.
The ramp video is fun, but i don't think it belongs in the same bin as other "optical illusions".
Yes, visual cortex failed, and it failed where a god would succeed (succeed in finding the simplest interpretation and making a sharp bet on that interpretation). But the most surprising optical illusions present a far lower bar which visual cortex then still fails to clear.
Often, they exploit what must be some specific implementation detail of visual cortex.
Often, they feel really weird from the inside. Sometimes they force competing interpretations, sometimes a single interpretation which turns out to be obviously wrong.
So i guess it's probably right to insist that an inference engine's optical illusions be a subset of its koans. At my level, the ramps are no koan, just a (funny and clever) lie.
2 comments
[ 3429 ms ] story [ 1762 ms ] threadI love the magnet like slopes - it's one that just makes you go "wow". Reminds me somewhat of the James Dyson "Wrong Garden" illusion as discussed previously on HN (http://news.ycombinator.com/item?id=779555, comments of thread discuss how to see a short video of it).
The "counter-intuitive illusory contours" one is interesting. It is almost as though our brain is treating the image as a liquid, and imagining distortion waves ahead of the "collision point".
The sine wave one is interesting in that you no longer notice the effect if you are reading the text below it (i.e not focusing on the wave beyond your periphery). This is somewhat similar to the "spinning dancer" illusion (http://en.wikipedia.org/wiki/The_Spinning_Dancer) in which you can trick your mind to see the illusion in different ways, and sometimes have difficulty seeing it again as you did originally.
In my related searches, found this site: http://yloveillusions.com/en/category/uncategorized/page/5/ which has some other interesting illusions. Particularly like the "A bottle disappears" one.
Yes, visual cortex failed, and it failed where a god would succeed (succeed in finding the simplest interpretation and making a sharp bet on that interpretation). But the most surprising optical illusions present a far lower bar which visual cortex then still fails to clear.
Often, they exploit what must be some specific implementation detail of visual cortex.
Often, they feel really weird from the inside. Sometimes they force competing interpretations, sometimes a single interpretation which turns out to be obviously wrong.
So i guess it's probably right to insist that an inference engine's optical illusions be a subset of its koans. At my level, the ramps are no koan, just a (funny and clever) lie.