Developing a demonstrator to showcase CEA’s AI-robustness assessment technology.
For CEA-List, we turned a tool for assessing neural networks into an interactive demonstrator that prospects can explore through software and a physical controller.
Context
CEA-List researchers developed PyRAT to assess how reliably a neural network responds under defined conditions.
That question matters when AI is considered for critical uses, from aircraft collision avoidance to medical diagnosis. But the mathematical processes behind the assessment are difficult to see, let alone explain to people outside the research team.
CEA-List wanted a demonstrator for its Paris-Saclay showroom that could engage researchers, industrial partners and other visitors without losing the substance of the science.
Mission
We worked with the researchers to decide what prospects needed to understand and how the analysis could be represented visually.
From that foundation, we developed the project identity, graphic system and interactive interface.
Prospects can change analysis parameters and follow the mathematical operations behind PyRAT’s assessment. We also designed and developed a dedicated hardware controller, making the interaction physical and giving the software a presence in the showroom.
Outcome
The demonstrator gives publics a way into a complex research subject: they can adjust parameters, observe the analysis and connect what they see to the question PyRAT is built to address.
Displayed in CEA’s showrooms, the interface and controller work as one experience. The project gave the laboratory a way to present its research through interaction, with enough depth to invite questions and raise interest of potential clients.
