(Attrib:Flickr/ ssoosay)
Artificial intelligence is one of those terms that has experienced a degree of semantic slippage over the years. Originally, it referred to a machine that had an intelligence that the average person on the street would recognize as intelligent. In time, the various parts that would go to make up such a system: expert systems, intelligent decision-making algorithms, chess playing computers, and even Jeopardy champions have been given the honor.
The most recent indication of the erosion of meaning happened fairly recently, when it was announced that Google had successfully trained a neural network consisting of 1000 machines to recognize cats in videos. Quite the accomplishment for a machine, but hardly very useful. However, it is an indication of the way things are going in artificial intelligence research.
Neural networks of the sort Google are developing require tremendous amounts of data to train them, and data is Google’s speciality. Their advances in Big Data technology for storing and processing enormous quantities of information may well be the key to developing much more intelligent systems than were previously possible.
We already see hints of how this can work with Google Translate, which uses the enormous quantities of data that Google has collected, particularly from the United Nations and other international organizations, to produce fairly accurate translations without human intervention.
The next step is to leverage that data and neural networks for purposes like voice recognition, object recognition, and improved machine translation. Google have vast repositories of video and textual information that can be fed to their neural networks, which are known as deep belief networks. A likely consumer application for this sort of technology will be in mobile devices, and particularly in Google’s much anticipated Project Glass.
In business the possible applications are very exciting. We discussed one benefit of intelligent decision making systems in last week’s post, but the possible uses go far beyond content discovery to intelligent trading systems for finance, vastly more efficient logistics, especially when coupled with the internet of things, improved market analytics, and many more.
Related articles
