An EU document compares machine learning with the invention of electricity. A total of 20 billion euros is to be invested in research into “AI made in Europe”.
A “Coordinated Plan on Artificial Intelligence” of the European Union envisages the increased use of algorithms in the areas of “migration, infrastructure monitoring”. This is the message in the annex to the communication from the EU Commission, which the Secretary General addressed to the Council shortly before Christmas. AI-based machine learning is to be used primarily in the areas of geoinformation and earth observation.
The EU operates the “Copernicus” programme, which initially consists of six optical and radar-based satellites. The images and geodata generated from space are used for environmental and safety purposes. Frontex, which requests satellite data for its EUROSUR border surveillance system via “Copernicus”, is regarded as the most important customer in the security sector. The EU Border Agency also uses satellite data to monitor the “pre-frontier area”. According to Frontex, EUROSUR is already able to use algorithms to distinguish between suspicious and unsuspicious ships.
Contributing to “evidence-based policy making”
The Commission also mentions climate change, environmental protection, agriculture, urban development, disaster response, and cyber security as further areas of AI application for “policy implementation and monitoring”. In civil protection, AI will also be used for “evidence-based policy making”. This refers to applications for crisis detection, such as those used by the German Federal Foreign Office, among others. This involves processing publicly available data, for example from the Internet, but also from Ministry databases. The Ministry of Defence is also researching such procedures with IBM.
In the Commission’s paper, data is referred to as the “raw material” for AI. The new Regulation on the free movement of non-personal data will soon apply in the European Union. Subsequently, the Commission intends to standardise the formatting of large databases, in particular “machine-generated data”, and make them available in “common European data spaces”. These are mainly Earth observation data and information from Copernicus.
Supercomputers for training AIs
The “Coordinated Plan on Artificial Intelligence” is bursting with euphoric formulations. The Commission compares AI with the invention of electricity. Accordingly, the proposals for financing public and private measures are far-reaching. As a first step, ongoing investments under the Framework Programme for Research and Innovation “Horizon 2020” will be increased to EUR 1.5 billion annually. Member States and the “private sector” are invited to “make similar efforts” in a public-private partnership. Over the next two years, this should bring together an investment volume of over EUR 20 billion. This is intended to expand, for example, “satellite technologies”, but also 5G mobile networks, fibre optic networks and “next generation clouds”. In addition, next year the Commission intends to make funds available “for start-ups and innovators” in the areas of AI and blockchain.
The AI investments envisaged are significantly higher than those of previous collaborations with industry. For partnerships in robotics (“SPARC”) and for mass data processing (“Big Data Value Association”), where a total of EUR 4.4 billion will be spent from 2014 to 2020. The “Coordinated Plan on Artificial Intelligence” is intended to build on this, as high-performance computing capacities are required for applications in the field of AI. The “EuroHPC Joint Undertaking”, in which the EU Commission and the member states are setting up a network of supercomputers, is mentioned. According to the plans, these will be used to train the AI.
Use for law enforcement and hazard prevention
In some Member States, test sites for “AI made in Europe” will be developed, building on existing centres of excellence. For example, support will be given to cross-border 5G corridors for networked and autonomous driving, and “real scale experimentation of smart hospitals”. Pilot projects are also planned in the fields of energy, healthcare, manufacturing sectors, geoinformation and agriculture.
According to the recently published German AI strategy [https://www.bmbf.de/files/Nationale_KI-Strategie.pdf], self-learning algorithms could also be used in law enforcement and threat prevention, e.g. in predictive policing, the recognition of prohibited content on the Internet and profiling of persons by analysing social media. The Commission also says that AI can help “to better prevent, detect and investigate criminal activities and terrorism”. Money laundering and tax fraud are mentioned. The “Coordinated Plan on Artificial Intelligence” thus supports plans to expand the cross-border use of financial information. According to a proposed directive, data from central bank account registers are to be used not only to combat money laundering and terrorism, but also to prosecute serious crimes.
The “Coordinated Plan on Artificial Intelligence” was drafted by the Member States, Norway, Switzerland and the Commission. It cements a strategic framework for national AI strategies in Germany, Finland, France, Sweden and the UK. All other Member States are now called upon to also draw up a national AI strategy by mid-2019 and decide on the investments and implementation measures. However, these appeals in Commission communications are not binding .
Final report in March
The research, development and introduction of applications to AI sometimes have great legal and ethical significance. For example, the Commission does not exclude their use for autonomous weapon systems. It emphasises, however, that this can only be partial autonomisation and that decisions on the use of lethal force must ultimately be subject to human control.
A High Level Group on Artificial Intelligence has now published a draft of ethical guidelines on the use of AI. The final report with recommendations should be available in March. Particular attention will be paid to possible applications in the areas of medical diagnosis and treatment, autonomous driving, insurance premiums and law enforcement.