Ref. Ares(2018)4839409 - 20/09/2018
Ref. Ares(2018)4949387 - 26/09/2018
Minutes - meeting Vice-President Andrus Ansip and CEO Hans-
Christian Boos (Arago GmbH), 25 January 2018
Participants
Hans-Christian Boos, CEO, Arago GmbH
Andrus Ansip, Vice-President, European Commission
Marie Frenay, CAB-Ansip
CNECT
Discussion
Arago is a European company at the forefront of Artificial Intelligence (AI). Arago offers a
general artificial intelligence platform and its mission is to help "clients’ transformation into
future-proof AI-enabled enterprises and to empower them to unleash the potential of their
existing intelligence". Today, most companies offer narrow AI that can solve only pre-
determined problems. Globally there are only 3-4 companies that offer general AI that can
learn from its experience and develop new solutions without new algorithms.
Mr Boos is a computer scientist. He insisted AI is about interdisciplinary research, about
several parts of science (not one single discipline). He added that machine learning is only
one part of AI, it is important to fund not only one part but different aspects (what Google is
doing). Current research system does not that properly.
Mr. Boos said that AI enabled systems can run existing business processes with 20%
of current costs. He mentioned Lufthansa as an example that is digitising the fleet
maintenance and connecting all planes to AI managed service centre. This handles
automatically the work that is equivalent of 500 service agents in the manual world. AI
is therefore a leverage to digitise industry. He also mentioned individualised chicken
feeding system that has increased the number of chicken that fulfil the strict quality
criteria from 30% to 85%.
He insists the money and energy spared can then be investing in new ideas,
experimentation (embracing failure, taking risks). It is essential because which
company can now "throw money" into crazy ideas except platforms which have
minimum costs?
Mr. Boos urged Europe to move fast and proposed several measures to accelerate AI take-
up in companies:
o
Europe should build on knowledge from the existing expertise areas.
Creating new products and services in existing expertise areas costs only
20% of the amount that it would cost if starting in a completely new area. This
means that Europe has huge potential in digitising its strong business sectors
as newcomers must invest 5 times more.
o
Data should be made easy to share and easy to trade. Public sector should
show the way and make the data available in platforms that are easy to reach
and in format that is interoperable with the commonly used systems.
o
European approach to privacy is a foundation for democracy and for
good business. It encourages creating products and services "that
customers don't hate" and European companies should take full benefit out of
this.
o
Europe should find a model to "retire entire industries". This could mean
for example paying 10.000 mining workers who will lose jobs to teach
everything from their work to machines. This would save the knowledge even
though the business will die. The knowledge could be re-used by companies
that build mining machines to global markets.
o
Europe needs urgently legislation that allows machines to make
independent decisions. This is crucial for self-driving cars, medical
treatments and other application areas but would require that responsibilities
and liabilities are clearly defined.
Example of their first AI cancer treatment
which, within three months, could make better decisions than humans.
Other elements:
AI is sometimes the new name given to things we have been doing for a long time (e.g. data
analytics).
Collecting data is important for training machines + description of the world.
Discussion about accuracy of maps and the automative sector. Maps done by Google better
than those of the automative sector. Accuracy is related to the situational knowledge, 1.6
billion devices collecting information live, data continuously updated. Maps based on factual
data might have more details and are very OK for autonomous driving, but they cannot be as
accurate.
Today people want a car where they can use their phone, not the GPS.
Issue of insurance: automative sector does not want to take risks. They could
"buy"/handle risks/insurance, they don't, they won't to avoid. And then you have
Google
which bought shares of insurance companies.