Since the invention of the wheel and especially since the triumph of the steam engine, new technology has changed people’s lives. And there have always been fears that man will soon become superfluous. None of this has come to pass.
Many AI experts therefore consider the fears expressed to be unnecessary and stress that the self-learning algorithms will above all help people. For example, by taking on tiresome routine tasks or providing support for very complicated tasks.
But the boundaries are becoming increasingly blurred. Already today some chatbot can be hardly distinguished from a human conversation partner at the telephone hotline. But does the customer really need to know whether he is talking to a person or a machine?
And what if the algorithm decides independently on the credit application and rejects it? Should artificial intelligence decide on a life-threatening intervention in a patient without a doctor? And who assumes responsibility if an autonomous vehicle is involved in a traffic accident involving personal injury?
EU Commission develops KI guidelines
The EU Commission sees an urgent need for action on these and a number of other issues and in April 2019 presented AI guidelines that were developed by a 52-member expert commission. Forty pages are devoted to describing the tension between artificial intelligence and ethics and making recommendations.
Roberto Viola, Head of the European Commission’s Directorate-General for Communication Networks, Content and Technologies (DG CONNECT), comments about the reasons: “We want an AI that takes into account people’s rights and needs and have tried to make our guidelines as integrative as possible.”
These recommendations are not binding, but they should provide a framework for companies and users. It focuses on four basic principles in particular: Artificial intelligence should not jeopardize human autonomy, always remain explainable and act fairly. In addition, its use should prevent social harm.
But can these high demands really be met? The experts are optimistic and have drawn up a comprehensive checklist for participants involved in the development of AI. Among other things, the checklist deals with the following aspects:
- How does an AI system react in unexpected situations and unknown environments?
- Is it ensured that people do not rely too much on AI in their work?
- Have the data sets been examined to determine whether they contain personal information?
Currently, organisations from all EU countries can test an “evaluation list for the creation of trustworthy AI” online and propose changes. Publication is planned for the beginning of 2020.
Swiss study: When algorithms decide for us …
Until then, results of the interdisciplinary expert study “When algorithms decide for us: the challenges of artificial intelligence” from Switzerland will also be available. Commissioned by the TA-Swiss Foundation it will evaluate until autumn 2019 the opportunities and the risks of AI by concentrating on five main areas: the world of work, education, consumption, media and administration.
Researchers from the Digital Society Initiative of the University of Zurich (UZH), the Technology and Society Lab of the Swiss Federal Laboratories for Materials Testing and Research (Empa) and the Institute for Technology Assessment of the Austrian Academy of Sciences (ÖAW) are working together on this project and analysing the technological, legal and ethical aspects.
The key findings of the study are recommendations to Swiss policymakers on the regulation of AI applications and support for the positive use of AI, as well as scientific findings on the opportunities and risks of AI applications in these areas. Some of these recommendations will also be relevant to stakeholders at European and international level.
“AI raises socially relevant questions that go far beyond data protection, for example when algorithms replace human decisions, trade on the stock exchange or influence voter behaviour,” according to the project. “Looking ahead, issues such as human enhancement or the fusion of man and machine offer a foretaste of the way this technology might reshape our future.”