Artificial intelligence
Algorithms – How the EU wants to understand and regulate the network
The logo of OpenAI, the maker of ChatGPT, on a smartphone screen. photo
© Richard Drew/AP/dpa
Almost everyone knows ChatGPT by now. But for a long time, partly opaque algorithms have been determining how we perceive the world on Facebook or Google. A new EU center is to ensure transparency.
The European Union wants to open up the black box of opaque software decisions on the Internet and make online giants like Facebook or Tiktok safer for their users. In order to actually understand the mechanisms behind hate speech, disinformation campaigns or psychological harm to children, EU officials will have to dissect the heart and soul of the platforms in the future: How do the recommendation systems work? What algorithms are they based on? What are the social implications of all this? The European Center for Algorithm Transparency (ECAT), which opened in Seville on Tuesday, is expected to play a key role in this.
The set of rules designed to bring order to the “wild west” of the internet is the Digital Services Act (DSA). So far, the EU has relied primarily on voluntary commitments by the services in the fight against illegal content. Among other things, the DSA is now intended to ensure that platforms remove hate speech and counterfeit products from their sites more quickly. Particularly strict requirements should apply to the very large platforms. The EU Commission will determine who will be included based on the number of users in the coming days. Facebook, Instagram, Youtube or Twitter are likely to be among them. They all organize their services to a large extent via artificial intelligence (AI) – i.e. algorithms. And that’s where the ECAT comes into play.
Algorithms control perception of the world
AI has long been largely responsible for how users experience the internet: What song will we hear next? Which post are we seeing? What are the results of our search? Which product should we buy? Algorithms determine the perception of the world so significantly and are becoming more and more sophisticated. This is of great concern to some. They warn that the self-learning programs could become uncontrollable.
The text machine ChatGPT is currently attracting global attention. You give him a few keywords or ask him a question – the answer is a fully formulated text that often cannot be recognized as computer-generated. There are now programs that can create images based on just a few words. The risk that such technologies could be used for fake news is obvious. Several MEPs recently called for a global summit on the dangers of artificial intelligence.
But back to Seville in Spain, where ECAT scientists are working – like in a German office – to ensure that the American online giants comply with European rules. A team of 30 employees, including AI experts as well as data and social scientists, will advise the EU Commission on enforcing the DSA and exchange ideas with other experts. Ultimately, sovereignty over the rules does not lie in Seville, but in Brussels.
Risk assessment required
In the future, the platform giants will have to submit a risk assessment with a view to harmful content to the EU Commission once a year – including possible countermeasures. The first report is due by September. It could be about the question of whether the mental health of children is at risk. The reports are checked by the EU Commission, which the public – researchers or journalists – can also look at. The EU Commission may request further information. She can question developers, request details about the test environment of the platforms as well as the code of the algorithms. Those who do not follow the rules face penalties of up to six percent of global annual sales.
Part of the rules are also that companies must remove illegal content quickly if they are informed about it. It will be easier for users to report such content. But is the EU really up to date with this law? The Commission is self-confident. Renate Nikolay from the responsible General Directorate speaks of “hardcore duties”. There is a great expectation that the EU will not only be good at making rules, but will also bring changes. “The world is watching us,” Nikolay said on Tuesday. The EU is the first association in the world trying to comprehensively put Facebook and Co. in their place.
But what about everything that is currently and will be developed on the basis of algorithms and artificial intelligence? What about ChatGTP and autonomous cars? What about automatic hiring or credit scoring programs? Artificial intelligence decisions can cause great damage.
Here, too, the European Union wants to be a pioneer and set standards. Two years ago, the EU Commission proposed a new set of rules, the so-called Artificial Intelligence Act, which is currently being negotiated in Brussels. In view of ChatGPT and Co., Digital Minister Volker Wissing (FDP) just warned to hurry. Here, too, ECAT will help to better understand algorithms and their effects on society.