We are living in exciting times for Europe’s technology strategy. While Artificial Intelligence has taken an explosive growth in the world, Europe is preparing several strategic initiatives with the aim of putting itself firmly on the global AI map. I am referring to the initiatives listed below, and in this post, we will see why they are important and how they (should) fit together.
- On June 26, 2019, at the first Assembly of the European AI Alliance, The European Commission (EC) presented its Policy and Investment Recommendations on AI, with the conclusions of the work of the High-Level Expert Group on AI.
- Earlier this year, in April 2019, the EC presented its Ethics Guidelines for Trustworthy AI, formulated by the same Expert Group after a public consultation process (500+ replies) on an earlier draft.
- Recognizing the importance of data for AI and for Europe in general, the European Parliament on April 4, 2019 approved the new Open Data and Public Sector Information Directive to stimulate the reuse of data generated by public institutions, publicly-funded private institutions and by publicly-funded research.
- Recognizing the value of privately-held data for governmental decision making and Artificial Intelligence, the EC has established an Expert Group on Business to Government (B2G) data-sharing to come up with recommendations on how to scale up the reuse of privately-held data for the public interest (disclosure, the author is member of this Expert Group).
Those four initiatives definitely will help to strengthen Europe’s competitiveness in the AI and data field. AI is fueled by data which is needed for the training of algorithms. Moreover, more available data will enable governments to take better and quicker data-driven decisions. The US and China are especially strong in the B2C data area, but there is an opportunity for Europe in the B2B space.
The Policy and Investment Recommendations on AI will stimulate AI development in Europe by i) fostering the uptake of AI in the private and ii) the public sector; iii) supporting world-class research; iv) creating a European data infrastructure; v) skills & education; vi) governance & regulation; and vii) funding & investments.
The objective of the Guidelines for Trustworthy AI is to ensure that all AI and big data developments happens within legal, ethical and save boundaries, and serve human kind. Also on the 26th of June, the EC launched a pilot to test the concrete assessment questions included in the Guidelines using two questionnaires, one for developers & deployers and one for other stakeholders. The pilot finishes on the first of December, 2019.
The Open Data and Public Sector Information Directive will stimulate the use of published Open Data through Application Programming Interfaces (APIs). It will also make a list of high value public datasets with high commercial potential that will be freely and openly available across the EU, via APIs.
The B2G Expert Group’s mission is to find ways to increase the sharing of privately-held data for the public interest such as government decisions, official statistics, humanitarian catastrophes, etc. Currently, such sharing is happening at a small scale, and often only in pilot settings; there are few systems in production. The challenge is to grow those activities as currently the public interest is not taking full advantage of the opportunities that data & AI brings.
To discuss the newest kids on the block (the Policy and Investment Recommendations on AI, and the pilot for assessing trustworthiness), we particularly welcome the recommendations related to stimulating the uptake of AI in the private and public sector; the creation of a European data infrastructure; the focus on responsible use of AI; and the focus on skills & education.
Source: Policy and Investment recommendations for trustworthy AI
Stimulating AI in the private and public sector
AI is currently mostly used by larger companies, but much less by SMEs. However, the large majority of Europeans are employed by SMEs. It is therefore essential to increase the uptake of AI technology in smaller companies. AI in the public sector has also to be increased: “Deploying AI systems can help governments make better evidence-based policy-making decisions, deliver better services to individuals, groups and organisations by reducing internal costs, increasing programme effectiveness, and enhance quality”. It is important to form (Public-Private) Partnerships for key areas to take full advantage of the technology. Training and AI advice are of critical importance to make this happen.
Towards a European data infrastructure
Most current successful AI is based on supervised machine learning, which requires large amounts of data for the training of algorithms. The current state of B2B data in Europe is, however, fragmented in organizations, sectors and countries. While many hurdles will have to be taken, it is important to work on a European Data Infrastructure as a critical differential asset for Europe. This is not only a hardware and connectivity infrastructure but mostly a data infrastructure. We must bring together data sets from both public and private organizations across sectors and European countries. The Open Data Directive enables the public sector to work towards this, and the B2G data sharing expert group has to come up with ways to scale up sharing of privately-held data, considering both public and private interests. Public-Private partnerships around specific areas are probably key to start building this data infrastructure.
Responsible use of AI
While there are many opportunities with AI and Big Data, there are also risks, which for a company are mostly related to unintended, negative consequences of applying the technology. We are therefore happy to participate in the pilot for assessing trustworthy AI and bringing in our experience with implementing “Responsible AI by Design”. Training, education and awareness are of utmost important, and we have developed an online course to train our employees on topics such as fair, explainable, human-centric, privacy-preserving and secure AI.
However, before we can reap the benefits of those important initiatives, there are still several hurdles to overcome. Here, I want to mention just two of the many, but less often discussed, issues that Europe is facing for creating a European Data Infrastructure. Privacy is of course also a very important issue, but already discussed at length by many experts.
- One unsolved problem is how to incentivize companies to share their privately-held data at scale to create this data infrastructure. Apart from scale, also time is important, because Europe can’t wait for 10 years. One of the main instruments is to create sustainable ecosystems and business models around certain niche areas like health, manufacturing, automotive, and define clear win-win models for the private and public sector. And for certain situations, a sharing obligation might apply. The B2G Expert Group will come up with recommendations by the end of 2019.
- Another problem is related to how to make data sets interoperable, that is, how to combine data sets coming from different organizations, both from the public and private sector. Standardization is needed. The full value of data can only be reached through the combination of different data sets. While the sharing of privately-held data is yet to take off, and we are “on time” to define a standard, Open Data abounds but is currently fragmented at different governmental levels (EU, member states, Statistics Offices, regions, cities), and there is no standard for how to publish it. The Open Data Directive should play an important role here.
In conclusion, there are many opportunities with AI, but also many tough challenges before Europe can become really competitive in this area. Telefónica is looking forward to continue collaborating with the European Commission and other key stakeholders to foster the competitivity of trustworthy AI.
Click on the video to watch the first Assembly of the European AI Alliance, where The European Commission presented its Policy and Investment Recommendations on AI, with the conclusions of the work of the High-Level Expert Group on AI.