DARPA is scheduling a 4 year program that essentially will give computers the ability to teach themselves. This particular initiative will make it easier to program, as well. This will boost artificial intelligence by developing complex "brain-like" neural networks. Even though it will inevitably be compared to the human brain, DARPA is not modeling it after an actual human neural network. There are so many things that we still don't understand about our own brains that it would be difficult to create a true artificial version of it. One of the main problems with this is each application will be extremely difficult to create. This has a lot to do with a lack of tools. DARPA also announced a new program to fund research in probabilistic programming languages.
Now these new innovations, if successful, will excite many people, but it will also scare the living shit out of others. I can't say that I'll be on one side of the spectrum or the other because like a lot of new scientific innovations there will be positive outcomes and drawbacks. As the article stated, the process has already yielded benefits. Self-taught computers will be able to complete very complex tasks and in some cases they will even be able to teach us how to do something that we can't even comprehend. This could yield a huge boost in productivity. On the other hand, if a detrimental problem arises what if we can't fix it? Let me explain. DARPA is essentially hoping that computers will be able to evolve on their own right? Well what happens when a program evolves beyond our understanding? If a virus is buried deep in layers and layers of code that we don't understand then how are we supposed to fix it? There seem to be a lot of potential risks. This is still years away from completion so only time will tell if this will be ultimately a huge positive innovation or a colossal mistake.
No comments:
Post a Comment