The topic of data science and machine learning seems to be off the headlines for some time now. We probably have made a habit of their awesomeness as they have become part and parcel of every business and every life. However, the developers and researchers are never at rest; and the applications of technologies like machine learning and deep learning have been silently growing smarter and more effective. And you know it if you have been keeping an eye on things.
A smarter personal assistant for you
One does, at times, come across some amazing stuff while endlessly scrolling through the Facebook feed. A lot of you may have seen Sundar Pichai’s presentation on Google duplex. If you have not, I will walk you through it.
Google’s assistant has taken its natural language understanding to a whole new level of efficiency with Google duplex. The duplex is a chat agent that can perform a number of tasks on your behalf. Suppose you need to book a table at a restaurant or book a session at a salon, the Google duplex can call the place and book you a slot without a sweat. All you need to do is to give Google voice command. It can comfortably comprehend the human agent on the other side of the phone regardless of the perfection of speech on their part.
Addressing the issue of broken English
In Pichai’s demonstration, there was a call between the assistant and a lady at a restaurant. The lady was evidently not a native speaker of the English language. She even had some problems understanding the request. For instance, when the assistant said that he needed to book a table on the seventh of the month she understood it as a table for 7 people. The assistant instantly recognized the mistake and used simpler language. It even had the presence of mind to ask how much the wait time was if they went without booking a table. The whole thing is quite fascinating but what goes under is an extremely layered and complicated technology.
Familiarize yourself with RNN
The technology used in Google duplex is called RNN. It is a branch of deep learning that stands for Recurrent Neural Network. It is a network where the output of one layer is added to the next input. So it is capable of creating a sequence of inputs and outputs that are all connected to the first layer. When you stack these RNNs it becomes an extremely complicated engine that can process a chain of commands. It is evidently different and more complex than the traditional feed-forward neural network which works with a single layer.
The point of this article
Introducing you with Google duplex and RNN is only one of the agendas of this article. The real point is to let you know that the market has a severe scarcity of data science professionals who can handle and take forward this kind of technology. Deep learning, machine learning, and Artificial intelligence are all gifts of data, without which these technologies would hardly come out of the books into the real world. You can be an active participant in the evolution of technology. Start with a data science online course. Get yourself enrolled and join the exciting world of data scientists. Who can say that you won’t be the next Jurgen Schmidhuber? By the way, he’s one of the developers of RNN.