Deep Mind


Deep Mind

Published on Feb 04, 2020 by Ashok Kasilingam


LinkedIn reports Artificial Intelligence Specialists, Robotics Engineers and Data scientists as the top emerging jobs for 2020. These jobs have grown substantially in the last five years, and there will be an increasing demand for skill sets like Machine learning, Deep Learning, TensorFlow, Python, Natural Language Processing in 2020 and beyond.

Companies around the world are very excited about Deep Learning precisely because of it's ability to recognize a pattern, optimize for a specific outcome and make a decision and the fact that this ability can be applied to so many different kinds of everyday problems. Google's acquisition of U.K Artificial Intelligence (AI) startup DeepMind back in 2013 could very well be attributed to this fact.

Quite often, we hear big Internet companies continuously collect our data through their free products and many people atleast in North America are now aware of the privacy concerns associated with using these products. There should be no surprise that these giant Internet companies like Google, Amazon, Facebook, Microsoft, Alibaba, Tencent, Baidu continuously collect and hold humongous  amount of data about their users and their behaviors. But what these companies do with such massive data and how does this all relate to AI and specifically Deep Learning is something which is not very clear to many people. To understand this, we have to look back to history of AI and it's evolution.

Ever since it's inception more than half a century ago, AI has undergone a number of boom and bust cycles. Traditional approaches to AI included “Rule-based” (an approach to teach computers to think by encoding a series of logical rules) vs “Neural Networks” (an approach which tried reconstructing the human brain itself). As we can see rule based approach is bound/limited by the rules that had been mastered by human brain whereas neural network approach relies on intelligently correlating millions of sample data to approach a simple problem (identifying an object/face recognition as an example).

Neural Networks which are now rebranded as Deep Learning could easily outperform other AI models at a variety of tasks. It is this ability to recognize deeply buried patterns and correlations by connecting humongous data points to a desired outcome is what huge Internet companies are after. From Youtube recommending the next video that you may like to watch to Amazon suggesting what you may likely buy next are all just initial outcomes of this deep learning. But the capabilities of Deep Learning are endless.

Data is the new oil which powers these Internet companies.