This week as I was looking for articles, there was one particular article that caught my eye almost immediately. It is called, “Artificial fly brain can tell who’s who” I knew that I had to write about this article. This article was posted in mid October (October 18th to be exact), so it is actually fairly recent.
This article talks about how researchers at the University of Guelph and the University of Toronto have built a neural network that almost perfectly matches that of a fruit fly’s visual system, and it can even tell the difference between other flies and even re-identify them. They obtained this by combining the expertise knowledge of the biology of the common fruit fly and machine learning to produce a biologically-based algorithm.
The article then talks more about the biology of the fruit fly, and talks more about the computer program in the following paragraph. The article then concludes by talking more about the future of neural networks and AI.
I throughly enjoyed this article, and I truly believe that it was well worth the read, and I encourage others to also find the time to read this article. The part that I found the most interesting was that using this neural-network-machine-learning-based program, this “artificial fly” was able to identify other flies with a score of .75 or about 75%. They tested this by recording the fly for two whole days and then testing the program on the third day to see if it was in fact able to identify it. They also tested just the algorithm without the fly biology constraints, and this scored a .85 and .83. This is only slightly better than the program which is very good results. They also went on to compare it to human fly biologists, and they only scored a .08. Lastly, on top of all of these comparisons, they included that random chance would only score a .05. This is unbelievable in my opinion. The fact that a computer program scored that much higher than a human is truly insane. I think that this research is a huge step in the right direction for AI. After reading this article, I am much more interested in AI, and plan to continue to research the topic more.