28 Mar The Artificial Intelligence disruption: interview of Manuel Davy, founder of Vekia
After the Mobile First disruption, here is another one that will impact the business world and especially retail and Ecommerce: Artificial Intelligence (AI).
France is embracing Artificial Intelligence thanks to its tech professionals who now play a key role in our economy. Although Google, Apple, Facebook, Amazon, and Microsoft (GAFAM) have managed to attract many of these professionals with competitive wages, there are still many others who create everyday algorithms for institutions such as the INRIA (French Institute for Research in Computer Science and Automation), CNRS (French National Centre for Scientific Research) and AI start-ups with innovative solutions.
AI has existed since 1953 however the landscape has significantly changed recently: cloud platforms makes vast computing power available at low cost, new tools allow easy customers data capture, GAFAM share and monetize their in-house algorithms, and nowadays companies continuously store more and more data.
The combination of those different elements marks the start into a new era, the Machine Learning one. Algorithms are not fixed anymore but constantly changing, learning in order to sharpen data (for example to make it predictive).
The French Institute for Connected Commerce (ICC) interviewed one of its best experts on this matter: Manuel Davy, founder and Executive President of Vekia. Before becoming an entrepreneur and founding his company, Manuel obtained a PhD in Applied Mathematics, worked in a lab at Cambridge University and has been researcher for the CNRS and INRIA. Today, Vekia employs more than 7 data scientists and Manuel is part of the creation process of complex algorithms with his team.
ICC: Manuel, do you think there is today an Artificial Intelligence ecosystem?
Manuel: Yes and no. Today, we could split AI companies into 3 categories: the ones that communicate about AI but keep a traditional statistical approach; the ones that use Machine Learning and AI via the tools offered by GAFAM; and finally the ones that contribute (such as Vekia) to create new algorithms and have a new AI approach. Those ones are rare. There is a real buzz about AI at the moment but many players use it in its current state without playing a greater role in its evolution.
ICC: Manuel, could you please clarify the difference between Machine Learning and Deep Learning?
Manuel: Machine Learning is a discipline composed by different sort of algorithms that aim to give a machine the ability to learn and then become ‘’smarter’’. Deep Learning algorithms are tools from Machine Learning that are a recent and extended version of neurons network. Deep learning is now very powerful when we have to teach to a machine how to identify complex information, e.g. how to differentiate between the picture of a lion and one of a tiger, or how turn oral words into written ones. Deep Learning requires numerous examples of what we want to learn and recognize. It cannot be used if there are not many representative examples. This is the case in retail for sales prediction: though there are millions of transactions everyday, it is rare to see examples that combine promotions, products and a certain place. It requires one more physical modeling steps, especially because data used for learning are often inaccurate or false.
ICC: In the AI jungle, how does Vekia position itself?
Manuel: Vekia decided to work on a specific topic, the supply chain. As a result, we developed AI algorithms to verify data, forecast demand and calculate the best supply chain strategy. Among the AI jungle, we position ourselves as true AI contributors: the highly demanding level of performance in our area requires to always be one step ahead of AI users. This is why our 6 tech professionnals closely work together in order to always deliver better results. Finally AI contributes improving performance for a more efficient economy as well as reducing waste and pollution; we then become a service of general interest.
ICC: How do you picture AI evolution in retail in the next 5 years?
Manuel: AI will reach maturity, it will become accessible to a majority of companies, in particular thanks to GAFAM tools. However, like for any trend, there will be some disappointments. Indeed the strengths, weaknesses and limits of AI tools are not understoody yet. In addition, we cannot get easily high performance at the moment. There is a Machine Learning theorem that express this: the « No free lunch theorem » which means that there is no algorithm better than another to solve a given set of issues. To reach higher performance, it is necessary to specialize the algorithm and sharpen it. Data scientists have a bright future ahead of them!