Artificial Intelligence (AI) relies on data both to make its decisions and to learn how to improve them; the higher the quality - and quantity - of data that AI applications can access, theoretically the better the outputs. However, data use is not without risk or consequence; for this reason, the European Parliament’s Special Committee on Artificial Intelligence in a Digital Age (AIDA) recently held a public hearing on ‘AI and the Data Strategy’.
Opening the hearing, AIDA Vice-Chair, Finnish socialist MEP Miapetra Kumpula-Natri compared AI and data to that of an engine and energy source, saying “One way to think about the relation between data and artificial intelligence is that if AI algorithms are the engine, data is the fuel. Even if we have the finest engine, it is useless if we don’t have the necessary energy source”.
“One way to think about the relation between data and artificial intelligence is that if AI algorithms are the engine, data is the fuel. Even if we have the finest engine, it is useless if we don’t have the necessary energy source” AIDA Vice-Chair Miapetra Kumpula-Natri MEP
This, she argued, “is why data should be at the centre of our attention as we chart the path towards a European future for AI”. She pointed out that several Parliamentary committees and numerous MEPs had worked to produce a report on the data economy, which had done an excellent job in setting out the conditions to deliver the expected benefits of the EU Data Strategy.
She accepted that managing the data economy posed significant challenges. The first being accessing data in usable formats. Too often, she said, it was siloed within countries and companies and was not interoperable. Unleashing its potential would require a new mindset. However, “it is not simply a matter of opening the data floodgates - it needs to be handled delicately; within an ecosystem of trust with fair rules”.
Kilian Gross, Head of Unit at the European Commission’s DG CNECT, explained that - from the European Commission’s perspective, “the three vital ingredients for a successful European data economy are access to data, the infrastructure to leverage this and the skills to use it”.
He agreed that data applications and services also needed to be trustworthy - including for AI. This has been an ongoing consideration for the Commission in developing its strategy. Swedish EEP Group deputy Jörgen Warborn said he believed Europe would benefit from a strong partnership with the US in the field of AI and data, arguing that “it was important to remain as aligned as possible”.
However, Spanish Socialist MEP Ibán García Del Blanco thought, “the EU should not be waiting for, or be subservient to, anyone on AI – it’s the US that is ‘behind on its homework’”.
Opening the first panel session, Jeremy Rollison, Senior Director, Data Policy & Digital Inclusive Economy at Microsoft, agreed that the importance of data in AI could not be overstated. “However, true value arises from the insights and outcomes from processing for AI applications”. This was a conversation that needed to take place; greater access to data offers one way of closing the digital divide, but he argued, that demanded a change in mindset. The key he maintained was collaboration.
“The three vital ingredients for a successful European data economy are, access to data, the infrastructure to leverage this and the skills to use it” Kilian Gross, Head of Unit, DG CNECT
Founder and CEO of the Eticas Foundation, Gemma Galdon Clavell, said society had experienced a recent “explosion” of data-based applications, particularly those using human data. These had been driven by US-based technology companies looking to boost advertising revenues.
These AI applications are now being repurposed to determine who can access services such as social and health care. “Transposing this low-quality AI - that cannot account for complex human processes and optimises for profit - had created huge problems. AI in high-risk contexts needs to be better than that”.
Stefaan Verhulst, co-founder of The Governance Laboratory, said new technology has huge potential for improving how we govern. “However, we need to be proactive in providing data access to make that possible, while avoiding misuse. We also need to better understand the impact of data, by gathering “data about data” if we are to be truly informed”. From a civil society perspective, Walter Palmetshofer of The Open Knowledge Foundation, thought the Commission’s EU Data Strategy had valid ambition. However, he was concerned over the practicalities of gathering such data.
In the second panel, chaired by Polish EPP Group MEP Radosław Sikorski, Professor Luís Paulo Reis, from the University of Porto, argued that AI offered considerable opportunities for improving public administration, saying “The ability to process ‘big data’ could assist in decision making; however, it was vital that any approach retain the confidence of the public”.
Data law specialist Professor Louisa Specht-Riemenschneider felt that effective data access for research was vital, and that access rights and trust systems could make this feasible. “The Data Governance Act should be supplemented by sector-specific regulation where required for example in Health Data”, she explained.
According to Thomas Bolander, professor of AI at DTU Compute, computing power alone was not the answer. He explained that “The road to safe AI lies in socially intelligent algorithms with higher cognitive function. If AI cannot match the human social perspective, then we can’t guarantee the proposed solutions will be acceptable”.
The next key milestone for the EU’s Data Strategy is expected towards the end of the year, with the Commission’s eagerly anticipated proposals for a European ‘Data act’.