Deutsche Bank’s analyst warns about GDPR posing risks to AI development in Europe
As the development of AI systems based on machine learning heavily relies on access to vast datasets, GDPR may restrict this access and hamper the development of AI in Europe, says Deutsche Bank’s Kevin Koerner.
Less than a month after GDPR – the new data protection regulation, came into force across the European Union (EU), the effects for the business and the general public start to become clearer. The chorus of concerns about the potentially negative consequences of the implementation is growing, with Kevin Koerner, European Policy Research analyst at Deutsche Bank, warning about the impact GDPR might have on the development of artificial intelligence (AI) systems in Europe.
In his article, Mr Koerner notes that development of AI systems based on machine learning heavily relies on access to vast datasets in order to train and improve their algorithms. This gives countries like China with more than 700 million (mobile) internet users and (so far) rather lax privacy laws a huge competitive edge. In Europe, however, restrictions under the GDPR could put the AI industry into a disadvantageous position compared to international peers.
In particular, Mr Koerner warns, a right to transparency in automated decisions (e.g. regarding online loan-applications) gives AI developers a headache as for machine learning techniques based on deep neural networks, the decisions behind the algorithm are a black box – even to the developers – and are evolving over time. As the Center for Data Innovation argues, the regulatory requirements to make algorithmic decisions explainable could reduce the scope and accuracy of compliant algorithms. This could put at risk the EU’s ambition to catch up with the US and China in the global race for AI dominance and to the opposite, rather cause the continent to fall further behind.
On the brighter side, Mr Koerner says there are also ways to mitigate risks to Europe’s AI industry. For instance, GDPR only applies to personal data. Hence, Automated (AI-based) anonymisation and pseudonymisation of datasets could help to allow for AI-related research and algorithmic development without breaching privacy rights protected by the regulation. Indeed, there are increasing efforts to develop more interpretable AI, such as the “Explainable AI” program initiated by DARPA, the US Defense Research Agency, and the LIME (Local Interpretable Model-Agnostic Explanations) framework. These may indicate how to address the black box problem and the concerns regarding the GDPR’s right to explanation. AI could also help companies to comply with the GDPR’s requirements, for instance, by handling user requests and managing databases, thus rather supporting than suppressing innovation.
Let’s recall that in April this year, the European Commission published a statement with regard to what it dubs a “European approach” to AI research and development but that announcement was pretty vague as it outlined abstract future plans and imposed financial obligations on the EU public and private sectors, instead of saying what the Commission commits to do on its own.
Back then, the Commission told the EU (public and private sectors) to hike investments in AI research and innovation by at least €20 billion between “now and the end of 2020”. No names of particular entities that are set to provide this pretty lump sum of money are mentioned. The Commission itself committed to increase its investment to €1.5 billion for the period 2018-2020 under the Horizon 2020 research and innovation program.
The rest of the announcement, unfortunately, contained nothing more than idealistic statements, such as a forecast that as a result of the growing adoption of AI solutions, “many jobs will be created, but others will disappear and most will be transformed”. There was no specification on which business areas are set to be affected by this trend.