Thought Leaders in Artificial Intelligence: Michel Morvan, Co-Founder of Cosmo Tech (Part 5)

Sramana Mitra: Let me try to explain what I’m looking for a bit differently. At some level, the discussion we are having feels a bit black boxy to me. What I’m trying to do is lift the hood of the black box and try to understand a bit more about what’s going on internally.

Michel Morvan: We created a computer language that is made to be able to describe this system. In this language, we created how to represent a system, a subsystem, and different types of interactions or couplings that you can have. For example, one system has its own dynamics. We have the language and the technology to represent the dynamics.

There is another system running parallel on the same big system having a different dynamics. Each one influences the other. First, you need a language to be able to describe these two different systems. This is what we created. You need to be able to describe the interaction. There are different ways to do it. We had a lot of constraints. In such a complex system, the interaction itself is part of the system. It’s an element of the system.

The way that two systems interact is part of the system. It can change in time. It can be influenced by other parts of the system. It’s not only the different parts of the system that you have to model, but it’s also the interactions themselves that are also elements of the system and you have to model the dynamics of that. We had to create a language and all the algorithms behind that allow us to create systems. You can couple systems at different levels, so the coupling, itself, is part of the system. The interactions also evolve in time. This is not the only thing we had to do.

We also implemented another thing. In all this subsystem, some of them can evolve very fast and some of them can have a slower evolution. We need to add the possibility for each subsystem and each sub-subsystem to be able to describe the way time evolves. It can be completely different from one system to another. Some systems can evolve every day or every second. Other systems can evolve only every year.

Sramana Mitra: One question that comes to my mind when you’re describing the system and this language – is your language different from domain to domain or do you have the same language tackling different domains?

Michel Morvan: Same language. That was a big challenge. When I created that with my team, I created that for  the biology domain. My main idea was with the same language, it should be possible to describe any kind of complex system.

Sramana Mitra: How many domains have you applied your language to and actually delivered complex decision making value?

Michel Morvan: Unfortunately, because we are a startup and we are only 75 people, we cannot go through all the domains that we can dream of. We apply it to biology. Then, we applied to cities. We applied it to energy networks, to the spread of epidemics, and to what happens inside of an individual. We also applied to a transportation system. These are different kinds of systems we applied to. We applied that to supply chain and industry. Every time, it’s the same language that we use.

We create software. Every time we create a model, we keep it. Now we have a library of models. When we create a new software solution, we can reuse a model. All of them are all on the same platform. One of the big powers of the platform is that when you create a new software solution, you have to write code. We reuse it so that the only code you have to write is for the way a subsystem evolves.

At the end of the day if you produce 100,000 lines of code, you have to write only 10% of that. You automatically generate 100,000 lines of code. It’s exactly the same platform. You can do it very fast because of the platform itself and all of the code that is already on the platform.

Sramana Mitra: What else should have I asked you about how you do things that I have not asked you?

Michel Morvan: That’s a meta-question. One question I would have loved is, “How long does it take to create a solution dedicated to a customer?” When we start working with a customer, the first solution we can provide to the customer is something that usually takes between three and four months. We can have very rapid results.

The second important thing is because we have encapsulated the expertise, we need some data, but we don’t need a lot of data to learn. We have the behavior because it comes from the brain of experts. The need of data is lower. Even if the customer doesn’t have all the data, it’s not a very deep problem.

Sramana Mitra: Great. Thank you for your time.

This segment is part 5 in the series : Thought Leaders in Artificial Intelligence: Michel Morvan, Co-Founder of Cosmo Tech
1 2 3 4 5