The third roundtable on ‘Building Trust in AI’ takes place on Tuesday, 22 October, from 11:00 – 13:00 at the Representation of the State of Hessen, rue Montoyer 21, followed by a light lunch. Here is the agenda or rather the set of questions to drive the discussion.
Standards and vertical contexts for trustworthy AI
For this roundtable discussion on creating trustworthiness in AI we would like to start laying the foundations of critical aspect such as standardization or prominent public policy issues such as cybersecurity and healthcare.
While AI covers a wide range of technology domains where many existing technical standards can be leveraged, advancements in AI technologies and applications are still in the stages of development and research.
- Does Europe need separate AI standards? Should Europe increase investment in the international standards bodies (ISO, IEC, ITU) for AI? What (if anything) needs to be done to foundational ICT security standards to ensure that they can be adopted for AI?
- Are there areas that might be more promising for pre-standardization work?
- What data-related standards would be helpful?
- Could there be technical standards for ethics?
AI changes the game for many specific vertical contexts. For instance, in cybersecurity, the typology of risks and attacks is different and grows in volume and complexity; on the other hand, AI helps predicting, analysing and mitigating cybersecurity threats.
- What are the typical software and hardware attacks on machine learning workloads?
- What are new forms of attacks e.g. adversarial attacks? how do they impact deep learning based systems?
- How can we apply security by design to AI systems?
- How can AI bring benefits for protecting critical infrastructure?
- What are key research areas combining AI and cybersecurity that need to be supported short term?
We would like also to start exploring other verticals (such as healthcare, finance, mobility etc).
- What are the advantages and risks linked to the use of AI in specific verticals?
- How does trustworthy AI appear in these contexts? Participants will be asked to share use cases examples.
- The European Commission is looking at the creation of European data spaces for improving data access and sharing. What are the pros and cons and how can the ecosystem support the process?