Wednesday, April 27, 2011

Autism - what lies beneath

I first heard about autism in a computer science class, Computer Processing of Pictorial Information, that I took at UMD. It's true that the professor, Yiannis Aloimonos, was quite unconventional and was telling us lots of stories about human and computer vision, and left us to decipher the equations used there by ourselves. One of his main ideas was that in order to make progress in computer vision was to understand as much as possible how the human brain processes the visual stimulus that results in human vision. But apart from recommending us books on human vision research he told us about a book of a different kind: clinical tales of a neurologist (The Man Who Mistook His Wife For A Hat). In this book Oliver Sacks describes various cases of conditions or accidents that lead to an abnormal functioning of the brain and how people cope in these situations. One of the chapters in this book is about autism. This is a born condition that affects some children and which impairs social interaction and communication (a less severe form is the Asperger syndrome). But a good number of these people show some remarkable abilities: superior skills in perception (e.g. the ability to draw very accurate texture details), attention and computing (e.g. the ability to tell in an instant is a fairly big number is prime or not). This lead to a theory that somehow there is a trade-off in our brain: the computing power there is sufficient to be able to do instant computation of prime numbers but most of the time is shut down and the brain is wired for things much more useful to us like the ability to communicate and interact socially with other people.
Now I'm sure you'll have a different view on some piece of news like this one: A twelve year old rewrites Einstein's theory of relativity.

Saturday, April 02, 2011

More than radio: semantic radio

Semantic web is basically the idea of adding some meta-information (usually tags) to the information presented on the web in order to facilitate automatic processing by hinting the computers about what that information is about. For example, the labels that I try to put to my posts that you can see of the left hand side of this blog. The big problem with this approach is the effort needed to tag all the information on the web, that's why there is also a big interest in processing unstructured data, which recently lead to a spectacular Jeopardy show won by a machine build by IBM.
Comming back to radio, this principle of meta data first occured a few years ago when appart from the audio signal the broadcasters added a station id meta data that was used by the radios in the cars to automatically discover the local frequency of an FM radio. This was a nice feature cause once you have your presets in your car then anywhere you would go the radio player will automatically identify the local frequency of that radio based on the radio id metadata.
But a few weeks ago one of my favourite radios (Radio Guerrilla) introduced an even more spectacular feature. They added metadata for the song currently played on the radio, so on the tiny screen in my car I could see the name of the song and singer while playing. I wouldn't care much about this feature on a classic rock radio, cause there I know most of the songs they play, but Gurrilla plays many new songs, especially British and some Romanians ones so before this metadata I was always trying to memorize some lyrics from the songs I liked in order to search for them and identify the song. Now I have right in front of me on the radio display. This will mean I will probably not know the lyrics of the songs anymore :)