Stéphane Roder is a French AI specialist. He studied machine learning at Stanford University with Luc Julia, the founder of Siri, and has deployed AI solutions in many companies. He has just published his Practical Guide to Artificial Intelligence in Businesses (Guide pratique de l’intelligence artificielle en entreprise Ed. Eyrolles, 2019). A guide in which he demystifies AI and gives it a more pragmatic future. The deployment of AI solutions in companies, he warns, is a necessity because AI is a revolution that affects all professions. We met with him.
DirectIndustry e-magazine: Why did you write this guide?
Stéphane Roder: There is an anxious climate surrounding AI today, with articles full of fears, talking about AI in China, how AI will replace humans… I worked in AI for several years. This gave me the ability to understand how AI solutions were developed and how they fit into a company. Today, the industry is not ready for it. Yet, it has much to gain from this revolution. That’s what I wanted to demonstrate in my book.
DI-emag: Are these concerns about AI not well-founded?
S.R.: In China, there is indeed a very intensive use of AI. But the use of AI in China is very much related to law enforcement, which is not our use at all. The Chinese have recommendation engines with a lot of data but they are designed to make recommendations for Chinese culture. They are absolutely not usable in the West.
DI-emag: What is our use in the West?
S.R.: Take for example the company Silex. This French start-up has developed a cognitive sourcing solution to help find a service provider. In short, we express in natural language what we want and with one click we obtain a list of results that corresponds exactly to what we are looking for. This is an example of how AI can be used for.
DI-emag: What is your definition of AI?
S.R.: AI copies man’s reasoning and understanding of his environment. Part of this understanding is statistical. AI has the ability to build a model from data and when we ask questions of the model it will give us answers. This is called linear regression or classification. All we’ve done in machine learning is classification. That means for example determining whether it’s A or not A.
DI-emag: Is this a revolution?
S.R.: Yes, because we are upgrading our information systems. For 30 years they were programmed. We are now integrating models into the programs that allow them to learn. They will make predictions, recommendations and actions. This ability to model environments is what is revolutionary. This is possible today because we have a computing power that we didn’t have before. We now know where to use each of the algorithms in the workflow, in the process.
DI-emag: What is the human’s place in this revolution?
S.R.: We’re going to take some unpleasant tasks away from the worker. For example, today payroll controls are done by hand by several people to avoid any errors. With AI, we will have assistance systems with people who are responsible for them.
DI-emag: But what if the human loses control?
S.R.: No one ever installs something they can’t control. There are always feedback loops. Everything we’ve always installed, we need to be able to uninstall it. We always check the quality of the response, the error rate. That’s the basis. It is the data scientist’s job to look at how the model behaves. Otherwise, in the industry this would represent considerable losses.
DI-emag: So the human is still superior to the machine today?
S.R.: Yes. And I would even go a step further. Why will there never be an autonomous car with today’s algorithms? Because we don’t know how to predict human behavior. Why did the Uber car hit a cyclist? Because it didn’t understand what the woman was going to do. Because human behaviour is not statistical. Today, we only know how to do statistics. So we will not have an autonomous category 5 car, i.e. completely autonomous. But we’re going to significantly increase car safety. It’s the assistance that’s going to be extraordinary.
DI-emag: What message should be sent to companies that are reluctant?
S.R.: All revolutions are scary. Man lives with his terrors, that’s the way it is. Some will exploit these fears and take advantage of them. AI is a huge step forward for our companies, and I fear that those who feed this fear will ruin this industrial revolution. If we don’t go, it’s going to be terrible. With AI, we will be able to increase the competitiveness of our companies. And those who don’t go will suffer from the competition of those who go. And this will happen within 3 years. Every time we put AI into a process in a very targeted way, we gain an efficiency of 10, 20 or 30% right away. It can be counted in the millions. When all this is understood, everyone will go. It would be like if today refusing to do office automation and wanting to keep typists.
DI-emag: But some jobs will disappear?
S.R.: What I see are mainly parts of jobs that will disappear. For example, bank lawyers. Each time a new financial product is released, it takes a lawyer 2 days to check that it is compliant, while it takes 30 seconds for AI. This will leave the lawyer more time to do consulting tasks. And we will need people to control the machine. So in a way, we’re going to bring back jobs. And then there will be new professions, around data. For the moment, we don’t really know how to how to exploit this data. We will need to set up data management, which will also create jobs. I am not worried about future generations. I’m more worried about the generation currently working. Those in their 40s-50s. How will they adapt? There will be a difficult transitional period to get through.. But any lawyer would prefer giving advice to searching overnight for the result of a judgment!