Letting engineering students prepare a class and teach it. This is the objective of the Artificial Intelligence master class launched last fall at engineering school ESILV. Clément Duhart, professor and researcher at ESILV and associate researcher at MIT Media Lab, created and launched this master class. He uses teaching methods inspired from the MIT’s.
The master class aims to share students’s knowledge and capitalize on it, thus enabling to stay aware of the evolutions of the vast and fast-moving sector of AI. Almost 90 students from year 1 to 5, participate voluntarily in the class, learning to work in autonomy, constantly exchanging skills. This is what Clément Duhart has experienced while researching in the USA. He wants to teach engineering students a more agile way of learning, by increasing their responsibilities in the class and letting them choose what and how they want to learn.
Here are some of the projects developed in the weekly master class.
Benefiting from the expertise of innovative companies
Clément Duhart invited Snips, a AI French startup specialised in voice recognition, to participate in the master class. Their embedded voice assistants can be implemented in a plethora of objects and applications : home automation, appliances, automotive, entertainment, retail. Snips’ technology is unique because the AI are on-device, which means that it is not relying on an Internet connection. Since it is not in hosted the cloud, the vocal assistants guarantee the user’s data privacy.
Snips team shared its expertise with two projects led by engineering students in the master class. A group of 3d year students designed glasses for visually-impaired people. A camera fixed on the glasses uses artificial intelligence for real-time identification of objects, reliefs, walls, etc and then relies on a vocal assistant to resynthetise the information.
Snips’ vocal assistant could improve the concept, enabling the user to ask questions aloud, for instance “is there an obstacle around ?”, instead of receiving a constant flow of information about the environment, whatever the user’s real needs. This project, created in the regular engineering curriculum after a company call for proposals has considerably improved in the master class. Snips’ help has been precious, as well as the ideas submitted by participants in the AI class.
“Snips will also work with student association DaVinciBot : they are creating a humanoid robot able to detect emotions. Students will use Snips’s technology to add a voice control system. Vocal interactions with the robots will then be possible”, explains Clément Duhart.
Using AI for ecology and music
Clément Duhart is part of the project Tidmarsh, led at MIT Media Lab. This Massachusetts wetland, an old cranberrie farm, is a life observatory where virtual reality and artificial intelligence serve ecology. Students taking part in the master class used Yolo, the real-time analysis and object detection technology developed by Støj, a Danish coding studio. They have used this system on ten cameras spread all across Tidmarsh to detect the animal species’s moves.
The MIT Media Lab is also actor of the master clas through a musical project. A group of students is working together with a PhD researcher from the Media Lab to map a musical texture on a midi tune. The system is able to track down all musical textures on any rhythmic and tune.
A virtuous circle for knowledge
The weekly 3 hour master classes start with a group of students giving a one hour class followed by 30 minutes of questions and answers. The following half hour are dedicated to the presentation of the practical work on the same theme. All students spend the rest of the class on the practical work. The following class starts by a report of the previous edition.
Students prepare and give the class. Clément Duhart only participates when needed. The month following the master class they will teach, students meet with Clément Duhart and ask for help on specific concepts. Groups count four students, most of them are responsible for as task.
Some prepare the slide and the speech of the master class, others are in charge of the practical work, including its coding. They use Jupyter, a notebook which issues a working document after the practical work and code have been typed in simultaneously.
Some students were appointed Teacher Assistants of Clément Duhart. The other students can rely on them for support in their projects.
The Artificial Intelligence master class aims to develop new innovative projects in the years to come, at ESILV and in the other schools of Pôle Léonard de Vinci, Business school EMLV and Internet and Multimedia school IIM.