Robust Artificial Intelligence
- Online -
11 - 14 January 2021
Venue: Online
If you are invited or already registered for this workshop, you have received login details by email.As industry moves towards more inclusion of AI in its products, there is a crucial need to investigate how academic and industrial research can contribute to the robustness assessment of AI in safety-critical systems. The present workshop aims at updating the community on the state-of-the-art in “Robustness in AI”, discussing how it aligns with industrial challenges and discuss a more collaborative future on AI safety between research institutes and industries.
It will specifically focus on:
-
AI robustness assessment against unexpected/adversarial perturbations
-
Formal and probabilistic methods for the verification of AI models.
-
Compatibility, adaptability and scalability of current techniques to industrial problematics, potential limitations
-
Ways forward towards their use in safety-critical systems.
Monday 11 January 2021
Start ~ 4:00 pm (CET time) - Time slots below show time since start of the day (00:00).
4h00 - Theme: Real-life applications - Industrial use cases
To follow this session, participants are invited to look at pre-recorded reviews on ‘industrial challenges’. All domain reviews will (ideally) follow a common format:
1- AI-driven use cases 2- challenges 3- current envisioned solutions 4- world-wide initiatives
| 00:00 | 00:10 | Welcomes by the Lorentz Center and organising committee | |
| 00:10 | 00:35 | Recap + Q&A on robustness of AI in the aeronautics domain | |
| 00:35 | 01:03 | Break | |
| 00:35 | 01:02 | Recap + Q&A: Law and AI | |
| 00:35 | 01:01 | Recap + Q&A on robustness of AI in the healthcare domain | |
| 02:00 | 04:00 | Break-out session: Theme: Aligned/diverging industrial AI robustness challenges. AI shortcomings. Short-, mid-, long-term solutions | |
| 02:00 | 02:30 | Introduction (panel) | |
| 02:30 | 03:00 | Break-out session / Brainstorming - Participants separated in discussion groups (x 3) | |
| 03:00 | 04:00 | Debriefing break-out session + panel discussion |
Tuesday 12 January 2021
Start ~ 4:00 pm (CET time) - Time slots below show time since start of the day (00:00).
4h00 - Theme: Robustness against unexpected or adversarial conditions
Empirical approaches to AI robustness assessment
To follow this session, participants are invited to view pre-recorded reviews/talks on ‘Empirical robustness’.
Reviews and talks are expected to go beyond the introduction of state-ot-the-art AI robustness techniques and focus on
their limitations and envisioned extensions and research ways forward.
| 00:00 | 01:00 | Flash talks recap + Q&A on pre-recorded reviews and talks | |
| 01:00 | 04:00 | Break-out session |
|
| 01:00 | 04:00 | Break-out session: Where lies the balance between AI architecture complexity and vulnerability to attacks? |
|
| 01:00 | 01:30 | Introduction (panel) | |
| 01:30 | 02:00 | Break-out session / Brainstorming - Participants separated in discussion groups (x 3) | |
| 02:00 | 02:15 | Break | |
| 02:15 | 03:15 | Debriefing break-out session + panel discussion | |
| 03:15 | 04:00 | Library SW demonstration: Adversarial ML |
Wednesday 13 January 2021
Start ~ 4:00 pm (CET time) - Time slots below show time since start of the day (00:00).
4h00 - Theme: Formal assessment of robustness
To follow this session, participants are invited to view pre-recorded reviews/talks on 'Formal robustness'
Reviews and talks are expected to go beyond the introduction of state-ot-the-art AI robustness techniques and focus on their limitations and envisioned extensions and research ways forward.
| 00:00 | 01:00 | Flash talks recap + Q&A on pre-recorded reviews and talks | |
| 01:00 | 04:00 | Break-out session Themes: Scalability - Complementarity of techniques? | |
| 01:00 | 01:30 | Introduction (panel) | |
| 01:30 | 02:00 | Break-out session / Brainstorming - Participants separated in discussion groups (x 3) | |
| 02:00 | 02:15 | Break | |
| 02:15 | 03:15 | Debriefing break-out session + panel discussion | |
| 03:15 | 04:00 | Library/SW demonstration: Formal methods #1 | |
| 03:15 | 04:00 | Library/SW demonstration: Formal methods #2 (parallel) |
Thursday 14 January 2021
Start ~ 4:00 pm (CET time) - Time slots below show time since start of the day (00:00).
2h30 - Discussions, conclusions & ways forward
| 00:00 | 01:00 | Panel discussion 1: Theme: "How will current and future research on robustness assessment techniques solve AI industrial challenges?" |
|
| 01:00 | 01:45 | Panel discussion 2 Theme: "Ways forward & research road map" |
|
| 01:45 | 02:00 | Panel discussion 3 Theme: "Future collaboration: Favored means & platforms?" |
|
| 02:00 | 02:30 | Panel discussion 4 Theme: "The workshop legacy" |
