Neural networking, a computational machine learning and pattern recognition technique inspired by the neurons of the human brain, has been bandied about since the 1940s, but a modern application of the technology in digital personal assistants has placed new attention on the decades-old technology.
This comes shortly after Microsoft divulged the results of some preliminary research involving neural networking, which states that the technology could increase speech recognition accuracy by about 25 percent.
In addition, experts maintain that neural networking has greater potential than current speech recognition technology.
It's no surprise that neural networking "is one of the hottest areas of development today," acknowledges wireless industry analyst Jeff Kagan. He mentions that digital personal assistants endowed with neural networking capabilities "will be the mainstream way we communicate with the world of information technology all around us."
Apple has been pulling in scientists from all over the world with expertise in neural networks to work on improving the functionality of its Siri personal assistant. But Apple is coming into the neural networking game a bit late, according to many analysts.
Already, IBM, Microsoft, and Google have deployed the deep learning technology in some of their products. Microsoft, for example, is using neural networks in the real-time translation feature that it plans to add to Skype later this year. Neural networking not only improves the accuracy of the speech recognition involved, but the translations get better the more the system is used. This is because the system is able to identify and learn from patterns in how people communicate across languages.
Microsoft's personal assistant, Cortana, also uses neural networking for improved speech recognition. Neural networking is a central component of IBM's Watson personal assistant technology. Additionally, Google has already implemented the technology in some of its Google Now applications.
In each case, the technology is being applied to learn about the user's preferences. Neural networking allows these assistant apps to learn about their users and become increasingly personalized, eventually carrying out tasks before the user has to ask.
Cortana, for example, stores and learns from personal information about the user, including likes and dislikes, interests, locations, personal contacts, and more, in her Notebook, a knowledge database that the system will rely on to deliver more personalized search results.
Neural networking also allows for multistep searches that create a conversational feel rather than a series of disjointed questions and answers. When asked to search for a restaurant, Cortana returns one result, based on the data she collected in her Notebook as well as Yelp reviews. Once the listing is displayed, the user can say, "Get directions there," and Cortana understands that "there" refers to the restaurant she just pulled up.
"The next level of personal assistants are those that are connected to us and help with many different networks," Kagan says. "This kind of technology will grow and let us use it in an increasing number of things we do every day."
Even on-demand Internet streaming media provider Netflix uses neural networking to compile taste profiles of subscribers based on their previous viewing habits. This information can then be used to make more informed recommendations.
Ray Wang, founder and chief analyst at Constellation Research, says neural networking technology could potentially lead to a more personalized user experience and yield solutions with much greater predictive capabilities. "Basically, we're moving to intent-driven systems that deliver mass personalization at scale," he says.
But, Wang points out that "the natural language processing algorithms require a lot of data points and a lot of time to self-learn and adjust."
Kagan has other concerns: "The downside is we will start relying too much on these assistants. Will we start to forget how to do things for ourselves? We just might," he warns.