EU Arti­fi­cial Intel­li­gence Whitepaper

Recently, a leaked doc­u­ment appeared online, which appeared to be a draft white­pa­per on Arti­fi­cial Intel­li­gence, pur­portedly by the European Com­mis­sion. The white paper sum­mar­izes the cur­rent EU approach towards assur­ing the block remains at the fore­front of AI devel­op­ment, and also prop­erly reg­u­lates future AI imple­ment­a­tions. The final white­pa­per is rumoured to be announced on Feb­ru­ary 19^th but, to give you a sneak peak, Data Insights read the leaked draft and sum­mar­ized a few of the key take-away messages.

The white­pa­per is excep­tion­ally per­tin­ent, as it con­cerns the cur­rent state of AI in industry. It also includes pre­dic­tions for future trends. For a con­sultancy firm as our own, the facts and expec­ted trends presen­ted therein will likely provide a strong help with eas­ing com­pan­ies into AI ori­ented solu­tions. For any­one curi­ous, here are a few excerpts.

• Cur­rently, in the EU, more than half of top man­u­fac­tur­ers have imple­men­ted at least one case of AI in man­u­fac­tur­ing oper­a­tions, com­pared to 30% in Japan and 28% in the US (Source: Cap­Gem­ini 2019).
• EU fund­ing for research and innov­a­tion for AI has gone up by 1.5b euros in the past 2 years, an increase of 70%.
• It is expec­ted that, in the next few years, less data will be stored in Data Centres, and more will be stored loc­ally at net­work edge-loc­a­tions. Cur­rently 80% of data is stored in data centres, and by 2025 it is expec­ted 80% will be stored at edge-loc­a­tions (factor­ies, hos­pit­als, etc.). Plat­forms will no longer be dom­in­ant in this area (Source: IDC Data Age 2025 study).
• In Decem­ber 2018 the Com­mis­sion presen­ted a Coordin­ated Plan – pre­pared with Mem­ber States – to foster the devel­op­ment and use of AI in Europe.
• “EU excel­lence” test­ing centres will be cre­ated that can fed­er­ate European, national, and private invest­ments. They will over­see AI invest­ment and research.
• The com­mis­sion and the European Invest­ment Fund will launch a pilot scheme of 100m euros in Q1 2020 to provide equity fin­an­cing for innov­at­ive devel­op­ments in AI. This will be scaled up by at least a factor of 10 from 2021 (through InvestEU).
• To ease com­pan­ies and indi­vidu­als into adopt­ing AI tech­no­lo­gies, the com­mis­sion set out an AI strategy on 25 April 2018 address­ing the socio-eco­nomic dimen­sions in par­al­lel with the rein­force­ment of invest­ment in research, innov­a­tion, and AI-capa­cit­ies across the EU (see COM(2018) 247). These are sum­mar­ized in the Guidelines on trust­worthy AI, pub­lished in April 2019.
• This has been tested by around 350 organizations.
• The white paper out­lines the defin­i­tion of ‘High-Risk AI’ as: 
  • sys­tem and products that endanger the human life, relate to pub­lic health, or present severe risks for the pub­lic order (for example, med­ical equipment)
  • national secur­ity (pre­dict­ive policing)
  • that involved in the pro­vi­sion of ser­vices within the pub­lic sec­tor, includ­ing pub­lic pro­cure­ment (for example in traffic man­age­ment systems)
  • that involved in the pro­vi­sion of ser­vices of pub­lic interest (air­ports, private hos­pit­als, and other ser­vices provided by private oper­ates with a pub­lic impact)
  • bio­met­ric iden­ti­fic­a­tion systems
  • recruit­ment processes

For these cases they list a num­ber of reg­u­lat­ory meas­ures that are expec­ted to be put in place, but have not been final­ized. These include train­ing only on European Data which meets stand­ards of qual­ity (e.g. non-biased data), model trace­ab­il­ity, trans­par­ency, and human oper­a­tional overseeing.

When the offi­cial white­pa­per is released, there will be con­tact inform­a­tion for AI experts, data spe­cial­ists, and pub­lic policy makers to send feed­back. Feed­back will be incor­por­ated before the sug­ges­ted reg­u­la­tions are turned into con­crete law. So, if you fall into one of these three areas of expert­ise, be sure to weigh in with your thoughts!