Sometimes, It Is Only A Matter Of Time |
Posted: April 9, 2019 |
Some ideas are just ahead of their time. It often happens that someone thinks up some brilliant new way of doing things and when they try to apply it practically it fails, not necessarily because it's a bad idea but because technology has not advanced enough to allow it to work. We saw this very recently in the area of artificial intelligence where expertise in machine learning is making great advances by using an idea originally developed in the 80s and then discarded. The Turing Award, which many believe to be equivalent to a Nobel Prize for Computer Science, was recently awarded to Yann LeCun, Geoffrey Hinton, and Yoshua Bengio for their work on Neural Networks. The idea behind Neural Networks goes back to the 1980’s when researchers got excited about the idea that computers could learn in the same way that a human brain does through exploration and experience of its surroundings and drawing its own conclusions about the world around it from the data. It became all the rage when it was first postulated and attracted a lot of people into the burgeoning field of artificial intelligence and machine learning. However when computer scientists tried to create machines that could teach themselves they found that the results were very mediocre and that other ways of trying to develop artificial intelligence were making greater advances and showed more promise. The idea became a bit of a backwater with mainstream AI research going in other directions. A few people keep exploring the idea and developing it, notably LeCun, Hinton, and Bengio, but most of the community moved on to other things. People who still saw the promise of the concept of Neural Networks were dismissed by many as wasting their time when other types of machine learning that were more more rule based and less intuitive were making real progress and attracting a lot of attention and research funding. It turns out that it was not the concept behind Neural Networks that was at fault, but the technology. Without advances in the processing speed of microprocessors as well as access to enormous volumes of data Neural Networks just learn too slowly. Now that the internet and the connected world creates reams of data every second and Moore's Law have made computing power inexpensive and fast it is possible to really see what Neural Networks are capable of achieving. It is now coming into its own and has become the most popular area of research in artificial intelligence and shows real promise at creating machines that can actually think and reason like humans. So don't be discouraged if some new brilliant idea that you have thought up turns out to be harder to implement than you originally thought. It may not mean that your concept is flawed. It could simply be the case that the technology needed to enable your idea to function effectively just has not been invented yet. It may only be a matter of time before the technology catches up enough for your idea to show its true brilliance.
|
||||||||||||||||
|