Neural Networks in AI

Neural Networks in AI: The Human Connection

All day long what we humans do is listen, taste, see, feel, and hear whatever by looking at what’s going around us, and then we react accordingly. Have you ever thought about what exactly is happening here? Your friend called your name and you immediately turned around to look at them, what did just happen here? The answer is Biological Neural Networks, yes the human brain the most complicated among all the species and is always at work with neurons firing all the time. Yes, you read it right! The “Neurons”1. The idea is very simple, all your five senses are gathering the data regarding our surroundings and feed it to our brain through our nervous system. From there on the relevant neurons fire in a specific combination by releasing a chemical through the “neural pathways” and the synapses 2.

Figure 1: A simple illustration of Bio-Neural Networks in Action

This is how and where the data is interpreted and made a sense of, like getting your hand closer to fire will get it burnt and until unless we never get it slightly burnt when we were kids, we never stopped trying. So that’s how the data works, once the child had their hand burnt in the fire, they now know that they need to maintain proper distance which now has become a pattern in their minds. Why are we discussing all this? Because we have managed to decode a part of the Biological Neural Network of human species and mimicked it in our computing systems, thus we developed the “Artificial Neural Networks” 3.

Neural Networks in AI: When Did It All Begin?

Usually we the humans of modern age believe that everything we have today in terms of technology is hardly a decade old which is not the case. Artificial Neural Networks isn’t that new at all, because it all started back in the year. Mathematician Walter Pitts and the neurophysiologist Warren McCulloch in the years 1943 portrayed it through a very simple electrical circuit and then Donald Hebb took it a little more seriously in his book The Organization of Behavior (1949) 4. The scientists took it to another level which changed the whole course of human history later was when such an Artificial Neural Network was translated to computational systems at MIT in the year 1954. Various other scientists kept on adding to it through the ’60s and then suddenly came the AI Winter. For a couple of decades, no significant research took place in Artificial Neural Networks at all, and everything was put in the repositories. In the 1980s the American Institute of Physics established and conducted the first-ever “Neural Networks in Computing” meeting precisely in 1987 5.

Figure 2: Donald Hebb, the man behind the mastery of ANNs

From there on, in the 1990s a lot of breakthroughs took place and to this very day it is exponentially developing, the only difference is that now Artificial Neural Networks are being deployed in various commercial fields like facial recognition, image processing, fraud detection, and many more. In this article, we will just take a look at the Artificial Neural Networks in a glimpse to get the basic ideas and then help the readers to get an interesting look for the next more technical one.

Neural Networks in AI: ANNs Explained Simply

Artificial Neural Networks are basically the subfield of Deep Learning that is an extension of Machine Learning which mimics human comprehension in algorithmic ways. ANNs work in a very simple way, they get the data (remember the data our five senses get?), find patterns among the data, and then train themselves on the identified patterns and finally make predictions on a new set of some similar data 6. We can very conveniently relate it to our example where a child learns about burning their hand in the fire, so a new set of data can be related to fire anywhere in the world as the child now knows that to avoid burning, they need to maintain distance. In short, Artificial Neural Networks simply mimic the human brain in order to solve data-driven complex problems. 

Figure 3: The ANNs rest with the Deep Learning in the AI order

Neural Networks in AI: Where Are the ANN’s Deployed?

  1. Data Classification: Ever heard about “Cat Recognition” through AI? It was done by feeding the data to the algorithm and once trained, utilizing it to predict. The same is used for facial recognition when the features of the face like face cut, jaw size & shape, eyes, forehead, etc are fed to the model.
  2. Speech Recognition: Imagine you and a friend are discussing a product with your smartphones put on the table and suddenly when you open your Facebook, the first ad that comes up is about the same product. Yes, Artificial Neural Networks can do it through Speech Recognition by learning the speech patterns.
  3. Time Series Analysis: Imagine you invest in stocks and also can get hold of the periodical data on the stocks as well, all you need is a trained ANN model that will predict the stock prices for you. How much will you make in stocks trade? A lot!
  4. Audio Generation: So, you like music a lot, what if the Artificial Neural Networks could do it for you? Just feed in the music as data and let the ANN model do the needful for you and produce a unique piece that soothes your ears.
  5. Character Recognition: You can train your ANN model to recognize your handwriting or your friend’s and also can help you verify your handwriting in case of forged signatures or statements. Impressive!
  6. Machine Translation: Can you recall “Google Translator”, they use Machine Translation to identify the chunks of text in one language to another and I just did this for you. In Spanish “ Sabemos que te encanta este blog” and in English “We know you love this blog”.
Figure 4: The Artificial Neural Networks, Very Similar to Figure one i.e. Biological Neural Networks

Neural Networks in AI: ANN in Action

Some of the major deployments of Artificial Neural Networks across the industries are detailed as under;

  1. Artificial Neural Networks are being deployed by the organizations to improve their marketing process and utilize their past data like customer preferences as well as shopping patterns to improve the market segmentation and thus targeting the most relevant customers.
  2. The pharmaceutical industry is also a big endeavor for Artificial Neural Networks and the biggest they have achieved is identifying the different types of cancerous cells such as IBM Watson Genomics is one of the major facilities to process large sets of data to achieve that 7.
  3. In the retail sector, Artificial Neural Networks are majorly deployed to forecast the stocks and enable the systems to ensure that all the products are available every time. Ocado the online grocers are one of the major utilizers using ANNs to predict their stocks in store in accordance with the customer demands 8.
  4. Cosmetics brand Sephora is using Artificial Neural Networks to segment the customers on the basis of various characteristics like age, spending patterns, location, history of products 9.

Neural Networks in AI: The Final words

There is no doubt that Artificial Intelligence is the talk of the town among all circles be it academics or across the industries. The idea is very simple, who so ever takes the advantage of the biggest resource they have i.e. data, and utilizes it to predict their future situation in any context through Artificial Neural Networks, will lead the coming age. In the fourth industrial revolution, ANNs will prove as the biggest and the most extravagant gateway towards Artificial General Intelligence. In this article, the main aim was to get our readers acquainted with this technology and help them to identify this in action at most common of places. I will come back with another informative blog soon, till then ciao!

1 Response

  1. September 28, 2020

    […] Want to read more on this topic? view our Artificial Neural Networks Article […]

Leave a Reply