How can we create trust in artificial intelligence and the European data economy? This was the question addressed at the Business Treff Trustworthy AI event at the beginning of October. Stefan Gindl from the Research Studio Data Science presented the RSA FG project TRUSTS.

Artificial intelligence (AI) has made rapid progress in recent years: Algorithms from this field now influence processes in industry, recognise patterns in medicine or select job applicants. The more decisions these systems make for us, the more important it becomes that they are based on “trustworthy AI”.

This was the topic of the Business Treff event “Trustworthy AI” organised by the Vienna Business Agency at the beginning of October. Stefan Gindl, Deputy Head of the Research Studio Data Science, presented the RSA FG project TRUSTS – Trusted Secure Data Sharing Space.

TRUSTS leads to trusted data markets in Europe

Since January 2020, the Research Studio Data Science has been working together with 17 partners from 9 countries on the international project “Horizon 2020 TRUSTS”. In order for Europe to keep up with big tech and global supply chains in the age of artificial intelligence, a functioning European data economy is essential. TRUSTS therefore aims to build a trustworthy European data market.

In “data rooms”, data from different partners is not to be stored centrally, but rather at the source and only shared when needed. TRUSTS wants to connect stakeholders from science, business and industry and create trust between them through certified services and security features. TRUSTS is thus closely linked to GAIA-X, the project to build a powerful and competitive, secure and trustworthy data infrastructure for Europe.

What do we need in order to trust AI?

In another keynote, Carina Zehetmaier talked about trustworthiness in her company Taxtastic. David Reichl from the EU Agency for Fundamental Rights gave insights into the legal and ethical difficulties of AI. In a subsequent exciting discussion round, the participants exchanged views on the challenges for AI.

A special focus was on the question of how AI systems can be made explainable and how people can remain in the decision-making loop – or whether they should. Another key topic was the bias of AI. Dealing wisely with biases in datasets can help mitigate existing stereotypes in society. At the same time, however, there are relevant biases that need to be preserved. A smart approach to this and similar questions can ensure that the risks of AI are minimised and the benefits outweigh them.