Already today, algorithms determine which
Someone advertising on Facebook is blurring or what prices are on Amazon
to show up. by artificial intelligence can also be systems in
Application processes a pre-selection or categorize customer requests
Determine which person receives a loan and which does not.
That doesn't have to be bad
because such technologies support people and make decisions faster
can meet. In the worst case, the systems can serve the daily lives of citizens
and citizens, because governments use artificial intelligence to fully control and control their citizens (like China
plans with the social credit system) or as decisions based on defects
Data is taken.
So how can you handle ethically
artificial intelligence? There are 52 experts on this question
on behalf of the EU Commission discussed for months. They introduced appropriate guidelines on Monday.
In the nearly 40 page & # 39; s paper they demand a reliable artificial intelligence whose operation does not violate any laws, which takes ethical principles into account and which is based on a sound technical and social framework.
Checklist for artificial intelligence
With their ethical guidelines, she wants to do it herself
the European Commission of China and the US. The two nations are currently engaged in a competition over who will take global leadership in the development and use of artificial intelligence. The European Union also sees enormous potential in technology. Europe must be present in the field of artificial intelligence, said Roberto Viola, director general of
Directorate-General for Communication Networks, Content and Technologies, before publication
the guidelines in Berlin. But you have to take advantage of the benefits
Be alert to the risks and challenges, Viola says. While in China the technology
previously controlled by the state and more driven by US companies,
Do you want artificial intelligence, rights and needs in the EU?
from the human. "We have tried to make our policy so inclusive
possible, "said the director general.
This statement reflects the difficult balance of the European Union: on the one hand, it wants to enable industry to develop artificial intelligence and to support it financially so as not to miss an important technological innovation. On the other hand, companies must follow certain rules so that the interests of private individuals are covered and the interests of civil society are safeguarded, In principle, the politicians hope to have a similar effect as they have achieved through the General Data Protection Regulation (DSGVO): that other countries look at Europe and the rules that apply there as a benchmark (even if the interest so far is at least the case of the GDPR is) mainly theoretical). "Europe stands for a brand and that means trust," said Viola.
According to the newspaper, companies, organizations or countries that develop artificial intelligent systems must adhere to four basic principles: artificial intelligence must respect human autonomy, avoid social damage, act fairly and be explained. These general principles make the experts concrete in further requirements for artificial intelligence aimed at people. For example, people should have the right not to make a decision fully automated if it puts them at a disadvantage. For example, when borrowing, this may mean that a customer has the right to talk to a bank clerk when their credit is declined.
In addition, the systems should not discriminate against people and should be developed preventively to minimize the risk of unintended consequences. For example, a company should test before use whether artificial intelligence discriminates or devalues certain groups of people. An example from Google illustrates how important this can be: the search engine image algorithm does not recognize black people as people, but they are classified as gorillas.
For example, EU experts recommend testing systems prior to use and considering the completeness of the data. Discrimination as in the case of Google could be reduced by various records, for example in which a company trains a system with photos of people around the world. In addition to further measures, the experts advise: a diverse team can therefore help prevent such mistakes, as well as social dialogue or training for both employees and citizens.