By Charity Delmus Alupo
According to this Forbes article, most engineers are scared of math. Unlike software development where the math is shielded from us through reusable libraries and functions, in machine learning, it is more directly applied. Yes, it is a popular opinion that math in machine learning is mandatory and there is no escape from it unlike in programming where an average programmer doesn’t get to deal with the mathematics on a day to day basis.
In an interview with Christophe Guéret of Accenture Labs in Ireland, he told his story transitioning from being a software engineer in training, to becoming Lead Research Scientist at Accenture Labs who are at the forefront of AI in the marketplace. He did his PhD in France, majoring in Information Dissemination in Networks. Christophe has also worked for the BBC as Senior Software Engineer amongst other exciting roles.
When asked what his hardest thing starting out was, Christophe admitted that it were the mathematical concepts behind machine learning as we know it. In his words, “I was not a hardcore formula person” and he did try to hide from those concepts. He was more interested in getting to see things work versus diving deep into what’s going on behind the scenes (for example how gradient descent works.)
But like many of us, it becomes apparent that the math is beneficial, especially when we can’t comprehend a dense research paper full of formulas, can’t reproduce a technique from a fine article, or more commonly simply failing to debug a line of code because it didn’t work as expected based on a tutorial.
Christophe had to overcome this fear and more, in order to get to where he is. I also asked him what attracted him to the field of Artificial Intelligence and he said “getting machines to learn to behave.” Interestingly, he never believed in Neural Networks which have now sort of become a synonym for machine learning for many but when you embark on a journey, you have to learn to adapt and grow.
Doing research in AI is very close to research in academia for so many reasons. One of them being managing expectations of a project. As Lead Research Scientist, Christophe said he has to deal with managing these expectations in a business setting when challenges such as requirement change requests or data related issues arise. His response to such has always been to be flexible. He also added that unlike in research where we have to push boundaries, in business it is okay to keep it simple. For example, sometimes all you need are quicker results and even though this might be less exciting, keeping it simple can be the answer. “It is not always about pushing the boundaries, but getting started,” he added.
My last question to Christophe was what the number 1 skill a machine learning PhD graduate should have. “Curiosity.” Always keep your eyes on multiple techniques, read papers you don’t usually read, interact with people, it does help a lot. If you focus on one specific thing, you will miss out.