Do we need more regulations regarding artificial intelligence (AI)? This question is being discussed internationally. Switzerland is also involved.
Legend:
More and more decisions depend on artificial intelligence.
KEYSTONE/DPA/A3634/Friso Gentsch
Thomas Schneider is Deputy Director of the Federal Office of Communications (OFCOM). As Chairman of the Committee on Artificial Intelligence in the Council of Europe, he is currently working on a European set of rules for the use of AI. In the “Talk of the Day” he explains what such laws could look like.
Thomas Schneider
Vice Director of the Federal Office of Communications
SRF News: Why is there a law for artificial intelligence?
Thomas Schneider: Because the technology opens up new opportunities in all areas. How, for example, the management of energy consumption or traffic can be made more efficient. If you know how many people take a certain bus at a certain time, you can adjust the offer.
But there are also risks.
But there are also risks. When the technology is misused for purposes that go against human rights or our societal goals.
There is already a legal basis, such as the EU General Data Protection Regulation or the revised Swiss Data Protection Act. Why are more laws needed?
This is true. The basic existing law, both national and international, applies regardless of which technology is used by the economy or the people. That means the Council of Europe Declaration of Human Rights, the right to freedom of expression, the right to an appeal, all these principles apply no matter how AI develops.
The various working groups asked: How far do the existing legal bases go? And where are there gaps that need to be closed?
Which laws are still missing?
For example, one gap is that it is not entirely clear how the right to a fair trial can be respected. There must be a possibility that an AI is contestable. With regard to criminal law, the question arises as to whether an AI may be at the end of a decision-making chain.
There must be a possibility that an AI is contestable.
Is it okay for a computer to decide on a sentence, or does that ultimate decision-making power have to rest with a human?
Data protection is repeatedly criticized. Is this concern justified?
The AI is a computational model that would not work without data. The goal is to allow AI to use data, but that people are always in control of their own data and have a say in how it is used.
Discrimination is an important topic in the context of artificial intelligence. Can a set of rules achieve anything on this issue?
There is actually this challenge that the algorithm itself can discriminate in an unlawful way. The problem is usually bad data that is not representative. In this case, the system further processes existing discrimination into new discrimination. Our goal is to find rules and regulations at both the data and algorithm levels that minimize this risk.
The conversation was conducted by David Karasek.