What is Neural Network and where is it used ?

We have seen artificial intelligence, deep learning, machine learning and we all know its capabilities and its uses. Now the question is - we already have everything needed to make a computer or any other device to become even more smarter than humans, then why do we need neural networks? Before answering this question first we need to know something. Do we really have everything needed to build a robot? of course yes but it is only 30% of what we are planning to build. Why do we build a robot ? to do things which we ask them to do and we would have programmed it in such way to do so. After building a robot have you ever given it a few options and made it to choose? That's where neural networks are used.

Imagine a situation where you had to choose between going to a movie on weekend or staying home. How would you choose between these two, you will start comparing it by giving priorities like you would first check your schedule on Saturday and confirm whether you are free or not, then you would think whether you would be happy by going to the movie or just staying home. You would start comparing the plan with various scenarios and will end up with a result. This is done by the neural networks on our brain. The moment you start comparing various scenarios to make a decision a neural network is created in our brain. This is the function of a neural network (MAKING DECISIONS).

What if the computers help us in making decisions when we are confused. To make that possible we  need to build a neural network. Building a neural network is not that simple, there is lot of programming and mathematical calculations involved in building a neural network. Allowing the computer to make a decision may sound weird but that will definitely be helpful if programmed successfully.

Google Duplex is a feature recently implemented into google assistant which can actually make calls when asked by its user to any store or even saloon etc. It can fix appointments for you and also reserve tables in hotels. This is done with the help of neural network. Duplex involves a lot of neural network.

At the beginning many claimed that allowing the google assistant to make calls from the owners phone may not be a good idea but later google announced that everything will be done with the users knowledge and not without their presence. That was pretty convincing but still allowing google assistant to make calls is still not such a good idea.

Back to topic - how google duplex works? it compares what the other person says over the phone and makes a decision accordingly. If it does not know what to do , then it will not make any further decisions and will just conclude with a negative reply to cut the call. For example :- if the mobile user asks the assistant to fix an appointment for a haircut at 12 pm on Monday, the assistant calls the saloon and asks for 12 pm appointment on Monday, if the person on the phone replies with a no then the assistant will ask for a time closer to 12 pm it can be either 11 am or 01 pm if both or not available and the appointment can be made only on Tuesday, then the assistant replies with a no and will inform you about the conversation. This can be done only with the help of neural networks.  We can build a software to make calls from our phone but the accurate talking can be made only with the help of neural networks.

From the above example you could have come to a conclusion about neural networks and its uses. If you are planning to build an assistant or a robot i would prefer to add neural networks to it, apart from being cool it will definitely be useful.