Regarding creative machines....

I've been scouring the internet for more information about Dr. Thaler and his so called "Creativity Machine". It turns out that he has a company called imagination engines, inc. that sells applications for his technology. I highly suggest checking out their homepage. There is some amazing information on it. This news article is also fascinating.

Apparently this guy has been doing things with this technology for quite some time now. You will be flabbergasted by what it has accomplished. Ever heard of the OralB Crossaction toothbrush? That toothbrush was invented by this machine after it learned from existing examples of toothbrush designs and their effectiveness. This technology has created music. In fact, as stated in the news article I linked to above, Dr Thaler has technically created more music than any composer in the world (of course it was his machine that did it, not him). The list of its creations, in fact just, goes on and on. There are endless applications for something like this. What seems most ironic is the fact that this technology invented itself in a way. You see, there are multiple neural networks involved in this technology. Some of the networks are like the trainees, and some are the trainers. One network is creating ideas while the other network evaluates these new ideas and rewards the good ones. Basically, this thing is self training! But get this...the segment of the network that evaluates the other network's ideas was created by one Thaler's neural networks. Thaler's second patent was created by his first...as in his first machine created the component of the second. It was like a team effort...the most revolutionary computing paradigm created to date (in my opinion) was developed by a team made up of a human and his created machine. Amazing.

If you want to see some of this technology for yourself I suggest downloading the screensaver that you can find under the product menu on the imagination engines web site. I have it running on my laptop as we speak and it is truly fascinating to watch. The web site stresses that this screensaver is NOT pre-scripted video....it is a real time rendering and the simulated robot IS actually learning during the screensaver's execution. It starts off with a simulated hexapod robot located in a maze-like room with a bunch of walls. When the program first starts running the robot is unable to walk effectively. It just flails its legs wildly about in an attempt to move. As time passes, it learns what movement patterns work effectively and which ones don't. This eventually leads to a coordinated tripod walking gait! Obviously this walking gait is familiar to anyone who has worked with this kind of bot before, but this bot actually learned how to do it on its own. Over time this simulated hexapod figures out how to turn when it runs into walls and even avoid walls to some extent with its sonar sensing. I can't imagine how well a full blown version of this AI would work with a robot or swarm of robots considering how impressive this toned down screensaver is!

I think Google should buy this company...with their computing resource and access to such large amounts of data they could probably make this technology do things that are absolutely insane. I bet there are endless applications for this technology in regards to search. Can you imagine how much better their video or image search engine would be if it didn't rely on filenames or user labelling (which people sometimes change to purposefully mislead) but instead used a creativity machine to tag each image automatically with relevant information? This would also be great for content filtering used by schools or parents.

Anyways, I wonder if any of the computer science department researchers here have experimented at all with technology like this? I would love to do research with someone here if they were doing that. I hope that I get to work with technology like this someday. For now I just want to keep reading more about it :)