Making Millions Out Of Artificial Intelligence: Brighterion CEO Akli Adjaoute (Part 3)

Sramana: How was the newly designed system received by the doctors?

Akli Adjaoute: It made a lot of news in France. For the first time there was a system that could react to data.

Sramana: Algorithmically, what were you doing?

Akli Adjaoute: For every application we did a smart agent. Object-oriented programming defines classes and attributes and then you define methods for the object. If you have a banking account, you can define a class as checking or savings with attributes such as debit or credit. The methods in object-oriented programming are where the programmer tells the objects what to do.

We define attributes to the object, but we also define goals of the object. If we had a soldier object, then its goal would be protection. We would define things it should do and should not do. In medical diagnostics, when a patient comes in they describe their pains to the doctor. In the mind of a doctor, he is thinking about a set of hypotheses. Then the doctor will confirm his hypotheses by asking more questions or doing more exams and tests. Depending on the results, the doctor can ask deeper questions or will arrive at a diagnosis.

In a computer, we do exactly the same thing. We take the patient inputs and list the hypotheses. Then we identified the treatment options for each hypothesis. Rather than referring to rules, we defined symptoms and associated treatments based on goals. For example, if the patient input is a fever, the system would identify that the treatment provided by Advil has a goal of reducing fever. We built a series of smart agents like this so that when a person came in they could describe a series of inputs such as fever, aches, and nausea.

Our system would then identify every type of disease and the inputs of each of those diseases. Rules are not needed at that point. Depending on what the person says, the system can intelligently match it to diseases that matched the symptoms. The goal of the smart agent was input-output just as a doctor had with the patient.

Sramana: What was the user model of this system?

Akli Adjaoute: My PhD work was for that particular hospital. We were able to solve some problems, just not at 100%. The system was able to propose the right diagnostic and treatment schedule without the rules. That was interesting because it was the first time that smart agents were reacting in a smart way without having a human right a series of rules.

Sramana: Is this the technology that you used to create your first company?

Akli Adjaoute: My first customer actually had nothing to do with the medical space. The first customer was a Belgium group that was looking for network troubleshooting. Every computer could then become a smart agent, and we designed goals for every agent on the network. We had a big screen where you could zoom in and see every computer on the network. Whenever a compute did not respond, its neighbors would query it and we could find out what the root of the network problems was. We reduced their troubleshooting from six to eight hours to one minute.

This segment is part 3 in the series : Making Millions Out Of Artificial Intelligence: Brighterion CEO Akli Adjaoute
1 2 3 4 5 6 7