It is very unlikely these days that you log in to your favourite social media platform or news aggregator service and not come across at least one article on Artificial Intelligence (AI). Viewing multiple articles on various applications of AI on any typical day is a routine thing now. Problems from every conceivable discipline are being tossed at a computer armed with a Deep neural network, and the results obtained end up amazing everyone in the field. AI systems have gotten better than us at certain tasks already, like identifying what an image is, in recognising speech, playing a few popular board games and recently even in successfully identifying anomalies from x-ray radiographs. There have been reports of AI systems that can write novels, songs, converse with a human, draw art to name a few. In addition, AI applications have also been reported in less obvious fields like journalism, design, and law.
The possibilities are endless. AI brings in the versatility and fatigue-less nature of computers into solving practical human problems. The AI-hype along with a few misconceptions (1, 2) in popular media may give us an impression that this is all just some kind of a magic gimmick. Let us try to understand a little bit about what underlies such systems and why they are being so widely researched and applied.
The pre-Deep learning era involved tedious research efforts towards, both input feature engineering and machine learning model (algorithm) design to solve a given problem. Hand-crafting features, as well as designing the learning algorithm, for a given task required in-depth domain knowledge and experience. Coming up with an algorithm step by step is a difficult and time-consuming process and optimizing its resource consumption is even harder!
One thing that lacked in such algorithms was that, they were not flexible and needed to be hard-coded as instructions specific to the problem at hand. Also, it was a major undertaking to adapt an existing algorithm to a new problem, which might just have a different set of attributes although of a similar nature. Once an algorithm is designed, a computer programmer would then code it into a computer step by step. When given an input to this computer program we would get an output, for example, given a certain set of health vitals, we get to know if a patient is healthy or not!
A Modern-AI (Deep learning) system, on the other hand, does away with “Hand-engineering” and reduces the need for expert knowledge to design the algorithm in order to solve a given problem. What it instead requires is that we provide it with large amounts of example inputs and their corresponding outputs and it can then learn to extract the appropriate features for a given problem and simultaneously use it to predict the output (1). 
So, for instance, if we measure the health vitals of many patients and we know who is fit and who is not and provide these pairs to an AI system, it will learn what the relationship between these vitals is and calculate the fitness of an individual. We therefore no longer have to give specific instructions, like say a particular vital has to be above a given range to be considered normal and so forth. Moreover, it can be easily adapted to a new related problem, using the initially learned model and new input-output pairs corresponding to the new problem.
This in effect makes AI more accessible as the difficult job of coming up with a method to solve a given problem is taken away. The other advantages like flexibility, robustness and improved results are few more reasons AI is being so widely tried out and adopted.
With such capabilities of AI, it’s possible that lot of tedious and routine industrial workflows will be handed to it along with various experimental tasks. We can certainly hope to have an AI-powered robotic assistant at homes and interact with them as part of our daily lives and soon become dependent on them. A small example of this is Amazon’s Alexa, which is a voice recognition AI assistant who answers the questions posed to it. The amazing part about such interactive AI systems is that they keep learning from us continually, during these interactions.
There has been a considerable amount of fear that the AI revolution will lead to a major loss of jobs and hence might be a bad idea. But in the opinion of experts this is not to be feared, since AI is a technology that will, in fact assist us in becoming more efficient and divert our focus towards exploring more challenging issues. The concern of job loss is a genuine one and has afflicted us at every major technology revolution, for instance, during the industrial revolution. Such a revolution can lead to advancement of society as a whole and take everything forward, but it might require the existing working population to pick up new skills, to adapt.
Although an important consideration that exists similar to the industrial revolution for instance, is the ethical use of this technology. This is a topic of intense debate across communities. How the technology is used is finally in the hands of the bearer. There was immense technological advancement during the Second World War, however the impact of the innovation depended on who possessed them. While, Adolf Hitler used them to cause havoc across Europe, there was also Alan Turing who nearly single-handedly stopped the war by cracking the Enigma code. Should ethics come in the way of innovating and developing an immensely promising technology? How AI is used and what impact it will have, is completely in our hands after all!
Here are a few blogs to get acquainted with the latest development in the field of AI:
There are many great podcasts (1, 2, 3) too, if that fits your taste.
Author
Saurabh Kumar is currently pursuing his Ph.D. at Department of Electrical Engineering, Indian Institute of Technology Bombay, India. Connect with him on Twitter @saurabhkm.
Editors
Roopsha Sengupta did her Ph.D. in the Institute of Molecular Pathology, Vienna and postdoctoral research at the University of Cambridge UK, specializing in the field of Epigenetics. Besides science and words, she enjoys spending time with children, doodling, and singing.
Paurvi Shinde has a Ph.D. in Biomedical Sciences (Immunology) with expertise in signaling pathways that boost T cell expansion/survival. She currently works as a Postdoc Fellow at Bloodworks Northwest in Seattle, where she studies immune pathways responsible for hemolytic transfusion reactions using transgenic mouse models that express certain human RBC antigens. Apart from science, she loves editing scientific articles, listening to podcasts and going on hikes in summer.
Illustrator
Smriti Srivastava says that apart from pipettes, colors are her buddies. She is a scientist by profession and an artist at heart. She likes illustrating science, sketching, landscape painting, Madhubani art form, traveling and photography.
Cover Image and inset images : Smriti Srivastava and Saurabh Kumar
Flow diagrams : Saurabh Kumar and Roopsha Sengupta
The contents of Club SciWri are the copyright of PhD Career Support Group for STEM PhDs {A US Non-Profit 501(c)3}, PhDCSG is an initiative of the alumni of the Indian Institute of Science, Bangalore. The primary aim of this group is to build a NETWORK among scientists, engineers and entrepreneurs).
This work by Club SciWri is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
