I started the design of a simulation engine with this in mind in fact. The idea being that, if the metadata surrounding whatever it might be that you want to simulate, were given the right structure definition tool, there would be no need for specialized simulation coding at all. All that we need to be provided would be the iterative loops required to go go through all of the specified, possible interactive characteristics. Those loops would then fire off SQL transactions (with their inherent advantages of already handling multi-threading, and memory management, and with the ability to do roll backs if required), if there were something to interact with, to a transaction server that could be load balanced across however many helper, sub servers, as the user might afford to throw at it. As such, the SQL database, as a whole, would represent the total interactional state of the environment of interaction capable objects.
I didn't take the project very far though. Both because I had gotten pretty burned out with programming in general, but also because, at the time, about the only thing I could see doing with it was trying to sell it as a piece of software, the process lost all interest to me. And now I wonder if I will ever get the bug to write software again.
DATASET WORTH 13 MILLION PICTURES
The Data That Transformed AI Research — And Possibly The World
Originally published in 2009 as a research poster stuck in the corner of a Miami Beach conference center, ImageNet quickly evolved into an annual competition to see which algorithms could identify objects in the dataset's images with the lowest error rate.
No comments:
Post a Comment