Source: Deep Learning on Medium
There is no doubt that the Artificial Intelligence is starting to make its presence in our day to day activities, for instance, suggested replies to an email, Alexa, Google Assistant, Siri etc. All of this is possible because of nearly 80+ years of research combined together with the booming and affordable computational power and more importantly the digitization of our society, resulting in the exabytes of daily data generation.
As a civilization, we have come very far and assuming we see the next 200 years[thanks to the global warming, alarming rise in population and scarcity of natural resources], AI will be almost everywhere. Whether it will utopian or dystopian society, it all depends on how our leaders will take decisions and how we choose our leaders [lets hope for better future leaders than what we have now].
Now lets focus on the need of the AI research. There has been so much of media hype about AI that it has created a sense of tension among some section of our society [thanks to movies like Terminator and Sci-fi authors for that]. To be honest, we are very very far from that situation.
AI is a very broad field and it has various sub-fields, one such is Machine Learning. There are various definitions of ML but recently, I read one [credit John V. Guttag, Professor MIT, Department of Electrical Engineering and Computer Science] which makes it very clear [at least from mathematical POV ], when Statistics meets Optimization, it results into ML.
Both of these fields are at-least 100 yrs old and have extremely rich mathematical theory and wide applications. With the availability of data and computational power they amalgamated and resulted into ML [Majority of ML problems are Optimization Problems but Statistics plays importance in the data selection, evaluation and validity of models]. But the point to notice is that all these were happening in various research institutes and research labs because in 60s-70s, most of the industries were focused on Industry 1.0 and 2.0 [electricity, mass production, mechanization etc.]
Fast forward to 2010, Machine Learning was at its peak among various researchers and the results proved for themselves. All the domain experts were using it and applying the various ML algorithms to solve their specific field related problems. These experts had the extremely important tool with them, Feature Engineering, which in very simple terms means what data to choose and what to reject for better performance of the algorithms. As a result researchers and scientists from various fields like physics, biology, finance, computer vision, speech etc. were enjoying the time of their life. Why? Because they devoted their life to be an expert in their field.
For the last 10 years, things changed drastically. Industry 3.0 resulted in the rise of internet and society adopted digitization. And researchers saw the rise of Deep Learning, a yet another sub-field of ML which is really making previously almost impossible things, possible [speech recognition, autonomous driving, object detection etc]. Deep Learning broke the barrier of domain expertise because it deals with the raw data itself, there was no need of Feature Engineering. But this was the result of researchers [Geoffrey Hinton aka Godfather of Deep Learning] collaborating with Google. Recently, big players like Facebook, Amazon, Microsoft have also entered this game with heavy investments, collaborating with or hiring the best AI researchers.
AI is a part of Industry 4.0 and can be driven only with the research mindset. Why? Because, first lets continue the past tradition and second its independent development and productionization is always possible. The focus should be on building more Algorithms [simple chain rule of differentiation was the inspiration for Backpropagation Algorithm, the brain of Deep Learning]
We are still far from achieving something extremely remarkable but that will be possible by investing in the research and collaboration between academia and industry, and last but not the least, supporting open source because none of this would have reached you without it [Google, Facebook and Amazon certainly agrees with it].
I hope this article helped you and in the next few articles I will try to cover technical content in more details.