Seetharaman Venkatraman, Senior VP, Innovation & Digital Transformation at Societe Generale GSC, pens down his experience at the NASSCOM Annual Technology Conclave which brought forth some of the biggest trends that are about to hit the industry.
As the flight descended amidst the dust laden smog in Delhi, the one thought that flashed my mind was how about letting loose a flurry of nano-drones into the atmosphere of Delhi. And how good it would be if these drones spray some kind of a pollutant-antidote (like spraying a Deo) all around at an altitude of 500m from the ground level. The antidote then absorbs all those harmful content from the air and through a chemical reaction crystallizes them into tiny harmless micro-water droplets that falls of gracefully on the earth surface. A guaranteed solution towards a pollution free crystal clear air in Delhi.
My thoughts were interrupted when I heard the announcement – “You have arrived at the Delhi Domestic Terminal”. Still lost in those hopeful thoughts, I left the airport towards the Conference Center that hosted the Nasscom’s Annual Technology Conclave for the year 2017.
Couple of us from Societe Generale GSC attended the event. It was quite intensive and immersive for tech-savvy enthusiasts. Lots of interesting insights. Presented below are a few glimpses from the event.
On the topic on Extremes of Machine Learning, Manik Varma, Senior Researcher at Microsoft presented good insights on “extreme classification”. He spoke about how collaborative filtering and matrix factorisation may not be accurate enough for predictions. He mentioned how MLRF or Multi Label Random Forest, Microsoft Edge ML Library (MEML) etc., could be well suited for use cases viz., for recognizing deep visual features from an image, Or for gene mapping / matching scenarios.
He discussed how very soon we would have models trained on the cloud first. And later the model could be ported to tiny IoT based micro-controllers with less than 2KB memory – thus paving the way for making predictions in an inevitable, connected-world of the future.
On another topic related to experiencing the future of work with Artificial Intelligence and Automation, Rohit Adlakha, VP and Global Head – Wipro Holmes, emphasized how customer expectations will evolve from Economics (Cost) to Experience and Efficiency. There was a reference to how AlphaGO (the first version of the AI game) was taught by humans (though supervised learning) and then it eventually defeated a world champion (human) at the ancient Chinese Game “Go”. And now we have AlphaGO Zero the latest version released in Oct 2017 which self learns by playing games against itself. Today its so powerful that if you pitch both AlphaGO zero and AlphaGo (the previous version trained by humans), the new version will defeat the previous AI Champion by 100 games to Zero. That sounded quite profound and well, I have to admit scary too.
On another discussion pertaining to data, the reference to quantum computing was made. The speaker went on to mention how incremental learning is better than batch learning (for ML problems) because in the former there is a steady flow of new data and weights get updated regularly thus reducing potential errors which are not the case in batch learning. Machine Learning enthusiasts – hope you have an interesting take away here.
The next session was on DevSecOps. The speaker, Pravesh Sharma, associate Director at Fidelity, spoke about how security is vital in the era of DevOps. He briefly discussed the various stages in a DevOp process and how Security can be integrated well into the overall DevOps process.
In the later half of the day there were many sessions and most of them were focused on AI. One of the speakers spoke about how sensors in vehicles could figure out which parts could go down based on usage pattern and data collected. It in turn alerts the vehicle service center to be ready with the spares (anticipating that the car or the vehicle could be brought into service soon enough). It would also trigger an alert to the manufacturer of the part to strengthen the engineering around the product – all of this in real time. WaW! There was a reference to Intels Xeon Phi processor which has the capacity to do massive parallel and vectorization processing for demanding computing applications.
We also got to hear how AI is used in Intelligent Talent Management from the CEO and CTO of Edge Networks. There was a session from Reliance Unlimit – a new IoT services venture from Reliance.
So in all it was an intense and enriching day to hear about Technology. As we headed back to the airport, I looked up at the sky with the hope to see if the smog could clear soon enough for an on-time departure of my flight back to Bangalore. And then came the announcement – “We apologize for the inconvenience. Your flight would now depart 2 hours later than the scheduled time”. Arghh! I felt like tumbling down a rabbit hole. What-if those nano-drones existed today! Want to join me in building one?