Future Mobility – Value of Data & Trust in AI

TDW II: Future Mobility
Value of Data & Trust in AI

Theme Development Workshop
28 October 2021 | 9:00- 17:30 CEST

Discuss with other experts the challenges and opportunities for AI in the Mobility & Transportation sector to jointly develop a Strategic Research and Innovation Roadmap for Trustworthy AI in Europe.

Programme

Due to Covid 19 the workshop will be held online with a mixed programme of presentations and in-depth discussions about specific sub-topics in smaller groups (Breakout sessions).

9:00-9:15Welcome & Objectives
9:15-9:45Trustworthy AI for Future Mobility / Dr. Müller – DFKI GmbH
9:45-10:15Safe, secure, and transparent AI: Thoughts out of an industrial perspective / Dr. Götz – ZF Group
10:15~10:30Coffee Break & Socialising
10:30-12:00Parallel Breakout sessions
12:00-13:00Plenary presentation of key findings from the Breakout sessions
13:00-14:00Lunch break & Socialising
14:00–14:30Explainability and Trust – The key acceptance factors of AI in automotive industry / Dr. Hilbert – VW Data:Lab Munich
14:30–16:00Parallel Breakout sessions
16:00-16:15Coffee Break & Socialising
16:15–17:00Plenary presentation of key findings from the Breakout sessions
17:00–17:30Closing & Socialising
Breakout session 1

Trustworthy AI for Future Mobility

With this session, we would like to initially define the strategic challenges and derive relevant major topics for Trustworthy AI from the perspective of future mobility.

Breakout session 2

Expl. AI for time series & verification approaches

Today Explainable AI approaches are used without knowing if they are appropriate for our data and model. This session will discuss the challenges of transferring theory to industrial applications due to theoretical limitations and lack of verification algorithms.

Breakout session 3

Estimating the value of data

The volume of data is increasing exponentially. But not all data has the same value for different stakeholders. To determine this value, a very efficient process is needed. This breakout session will therefore answer the following question: How can the value of data be estimated?

Breakout session 4

Towards standardisation & certification of AI

For non-machine-learning components, standard processes and methods exist to prove safety, but for machine-learning components, corresponding standardisation and validation processes still need to be developed and certified. This session will discuss the challenges in this context of camera sensor-based algorithms.

Breakout session 5

AI expertise in Future Mobility

The mobility & transportation sector faces several challenges in attracting talents and empowering their employees to provide AI-based solutions. What are the specific needs for AI training and upskilling programmes, and how can these needs be aligned with academic activities and doctoral programmes?

Breakout session 6

Data for Future Human-Centered AI applications

Data is collected for its information and to serve a specific purpose for the user. At the same time, it also contains information that can be ethically challenging. This breakout session will discuss the properties of data and how it connects to other ethical and trustworthy AI initiatives so that it can form the basis for future human-centered AI applications.

Breakout session 7

Reliable Confidence Measure

Currently used deep learning methods do not produce sensible confidence. Therefore, the reliability of deep learning methods in safety-critical use cases cannot be assessed and trusted. This problem is further exacerbated in the presence of adversarial attacks. In this breakout session, we will discuss reliable confidence measures to handle these challenges.

Breakout session 8

Machine Learning in the context of personal data and GDPR

This breakout session will discuss the possibility of enriching machine learning with personal data based on GDPR. In particular, the question of how to protect individuals but use the information to create human-centred AI applications is of importance in this context.

Breakout session 9

AI sensitivity analysis for time series

For the application of AI in safety-critical situations, it is important to know the influence of input signals on an output. In this session, a first analysis of existing time series data in automotive applications will be done to define safety levels, safety integrity levels and the safety criticality of the current situation.

We invite the community to suggest further topics of interest for the breakout sessions. Please use the online application form for your suggestions.

Theme Development Workshops cut across multiple ICT-48 Networks of AI Excellence Centres while bringing together researchers, industry representatives, and other stakeholders to identify industrial trends and needs and match these to AI capabilities in Europe.

TDWs are co-organised by the VISION project in close cooperation with the TAILOR and HumanE-AI-Net projects as well as with CLAIRE AISBL.

Previous TDWs

1. TDW: AI in the Public Sector

Organising Committee