Submitted for publication in a magazine for restricted circulation within the employing organisation, all rights reserved. ©
In the recent years, there have been phenomenal advances in the state of the art in Deep learning and Deep neural networks, one of the most promising directions for Artificial Intelligence. Most of you would have seen the Robot “Sophia” holding her own in conversations with humans. She (it?) was recently given citizenship in Saudi Arabia as well.
Less fancy but more useful applications have transformed our lives in many subtle but important ways.
- Spam Catchers
- Optical character Recognition
- Natural language processing (Think Google Assistant, Siri, Cortana or Alexa)
- Medicine and biology
- Detecting electricity theft (https://www.engadget.com/2017/09/25/companies-will-use-ai-to-stamp-out-electricity-theft/)
Neural networks are one of the most beautiful programming paradigms ever invented. In the conventional approach to programming, we tell the computer what to do, breaking big problems up into many small, precisely defined tasks that the computer can easily perform. By contrast, in a neural network we don’t tell the computer how to solve our problem. Instead, it learns from observational data, figuring out its own solution to the problem at hand. So, in essence, instead of writing detailed instructions for the computer, you show it the output you want and let it figure out the instructions by itself. Isn’t that how you teach a human?
The power of a deep neural network comes from the possibility to learn very complex non-linear functions. For example, the following shows the impact on revenue from consumer churn.
Automatically learning from data sounds promising. However, until 2006 we didn’t know how to train neural networks to surpass more traditional approaches, except for a few specialized problems. What changed in 2006 was the discovery of techniques for learning in so-called deep neural networks. These techniques are now known as deep learning. They’ve been developed further, and today deep neural networks and deep learning achieve outstanding performance on many important problems in computer vision, speech recognition, and natural language processing. They’re being deployed on a large scale by companies such as Google, Microsoft, and Facebook.
What is a deep neural network? It is the name we use for “stacked neural networks”; that is, networks composed of several layers. The layers are made of nodes. A node is just a place where computation happens, loosely patterned on a neuron in the human brain, which fires when it encounters sufficient stimuli.
Deep neural networks have surpassed humans in many image and language processing tasks already but they are still narrow intelligence rather than general intelligence. The frontier of research in the field is moving extremely fast. Some beautiful examples of generalization can already be seen (https://arxiv.org/abs/1705.03633 and https://cs.stanford.edu/people/karpathy/main.pdf
) but there is still a wide gap. The future is indeed exciting.