3/4 – You want to learn more about AI but you are a little intimidated by the subject? Don't worry, Romain Chaumais, executive leader of the start-up Fashion Data, specialized in artificial intelligence applied to fashion, tells you everything there is to know about it.
By Ludmilla Intravaia
Le Boudoir Numérique: In the first two parts of our interview (read here and there), we discussed, in a very concrete way, how artificial intelligence can be useful in fashion. Let’s close this discussion by looking more generally at what AI and its technologies are. But first of all, how did companies come to rely on artificial intelligence for their decision-making?
Romain Chaumais, executive leader of Fashion Data: Data processing has often made it possible to better understand one's activity in order to effectively improve decision-making, by obtaining the knowledge necessary to carry out good arbitrations. This approach, followed by companies for several decades now, is making increasing progress. The idea is to analyze more and more the information possessed to move from an intuitive individual judgment to a collective quantitative judgment. We first started with the observation, based on figures, in order to answer the question: What happened? From this inventory of what had just happened yesterday, the day before yesterday or last year, it was possible to deduce actions to be taken to improve and optimize the business. It’s business reporting. Then we came to an analytical phase to answer the question: Why and how did it happen? And so trying to figure out, for example, why one store performs better than another, why this item didn't sell as well as this one, why this customer stayed loyal and the other didn't, etc.
What is different with this analytical approach?
With analytics, we moved beyond just a few numbers on sales, inventory and customers, we started to need more data and more powerful tools to process it. At a technological level, the more we want to understand, the more we have to call on detailed data. This level of detail requires computing power, large storage capacity and specialized tools to be able to navigate these huge amounts of data, now generated by the data deluge. Added to this is data visualization, the graphic representation of data, because we realized that, faced with endless listings or tables of data, we could no longer understand the reality of the figures. Putting information into perspective, in a circular diagram, such as a pie chart or a bar graph, allows you to immediately identify the causes and consequences of a phenomenon in order to better understand it. Finally, over the past ten years, business intelligence, traditionally used by top decision-makers, has also become accessible to as many people as possible in the company. Today, in data-driven companies, everyone from the CEO to the manager of a warehouse can rely on information to optimize their decisions, thanks to data democracy.
What about artificial intelligence in all of this?
The mathematical concepts foreshadowing AI date back to the 1930s and its first computer models were developed in the 1970s. But a major obstacle to artificial intelligence remained the inability to achieve the computing power necessary for its development. However, thanks to cloud computing and faster and faster microprocessors, it is now a done deal, in a digital context of deluge data that has led to the appearance of new tools, so as not to be drowned in constantly increasing quantities of data.
Are these new tools what we call big data?
Big data is the technology developed following the data deluge that generated quantities of information that, still ten years ago, could not have been handled with traditional tools. You have to understand that a website produces 100 times, 1000 times more data, considerably more data per day than that relating to the sales of a large company, for example. Each individual who clicks on a website, each page viewed, each RFID chip scanned, each use of connected devices of the internet of things and smartphones, all this information is now collected, creating this data deluge.
Is this deluge an increase in the volume of data?
The number of data is now estimated to double every eighteen months. These data therefore increase in volume, but also in variety. It's not just tabular info like in an Excel spreadsheet. It can be text files, images, sound, etc. Their velocity has also increased. Ten years ago, you could update a company's information every week, or even once a month. At the beginning of 2000, a bank updated its data once a month. Today, once a day, that's the minimum. In the age of big data, no human being is able to analyze, alone, the growing data, in volume, variety and velocity that is collected.
Why ?
We are very good at intuition, at understanding a phenomenon as a whole. Much less to analyze, in detail, raw information to understand what has just happened. Our cognitive abilities are too limited, our brains are not made for these kinds of tasks, at least not on a large scale. Artificial intelligence, on the other hand, is capable of understanding immense amounts of information and of doing, on a large scale and at high speed, what the human brain does, on a smaller scale. This allows computer programs to analyze what happened, what is happening, and deduce what will happen, using machine learning.
What do you mean by that?
Take the example of image recognition. If I give you fifty pictures of chicks and hard-boiled eggs, asking you to separate the chicks from the hard-boiled eggs, you are going to be able to do it. If I give you a billion pictures, in which I ask you to find all the chicks, it will get very long and very tiring for you. AI, on the other hand, is able to perform this task through machine learning. You are going to tell it: this is a chick, this is not a chick and, in the end, it will understand, on its own, what a chick is. You may not know how it learned, but still, at some point it will have learned enough to be relevant to this topic.
More relevant than a human?
Yes. This is the reason why artificial intelligence is disrupting professions like radiology, for example, professions requiring interpretation skills because it is less mistaken than a human. Because it was able to learn from billions of cases, and not just thousands, like a human being. Artificial intelligence acts like a huge brain that has the time and the power to read everything, to understand everything, to give you a summary and then can predict what is going to happen.
* Continue to explore the world of artificial intelligence, next week, in the last part of Romain Chaumais’ interview: Data scientist, algorithm, computer vision… Let’s talk about artificial intelligence!
* Read the first part of Romain Chaumais' interview on Le Boudoir Numérique: "Smart data at the service of fashion eco-profitability". The second part is here : "Big data enables intelligent management of the fashion stores supply chain".
* Fashion data website is here.
* Continue reading on Le Boudoir Numérique with these following papers:
- What will we wear post-lockdown ? The answer with Heuritech’s AI
- Covid-19 – Heuritech launches a solidarity program for fashion brands
- “Cross-fertilization between tech and fashion is the strength of Heuritech”
- “To help fashion brands decision-making with AI during the Covid-19 crisis”
- Will leopard print pleated skirts be trending any time soon ? AI already knows it !
- “AI can contribute to the virtuous circle of sustainable fashion”
- “Our algorithm is the link between the parfumer and the customer”