Overview
We conceptualise Responsible AI (RAI) as the conscious effort in designing, developing, and deploying artificial intelligence (AI) systems in a way that maximises their benefits while minimising their risks. The past few years have been characterised by the rapid development of sophisticated AI models along with rising ethical and societal concerns. This leaves us with the difficult puzzle of optimising for performance without compromising core responsible AI principles such as, transparency, robustness, environmental sustainability, etc. This challenge builds upon the previous Carla Challenges. Added to the existing performance assessment metric, we implement new ways to also assess and quantify different responsible AI concepts such as robustness, sustainabiliy, and transparency.
Carla Responsible AI Challenge assesses the robustness, environmental sustainability (carbon emissions) of the submitted agents at inference time. See instructions on how to get started
The top performing entries across the different RAI category will receive invitations to present their work in a workshop whose details would be provided soon. In adition, all participants are encouraged to submit a short and concise technical report (maximum length of 4 pages) detailing their submissions. This should include methodology and any other useful information regarding transparency and sustainability choices or considerations.
Timeline
| Actions | Dates |
|---|---|
| Challenge Opening | April, 2024 |
| Challenge Closing | July, 2024 |
| Video submission for top teams | July, 2024 |
| Workshop for result presentations | TBC |