The popular ChatGPT program is sparking a discussion about Artificial Intelligence (AI) in Europe. This year, the EU will introduce AI legislation that should, among other things, prevent human rights violations. However, the system behind ChatGPT would so far fall outside the legislation.
ChatGPT is a self-learning computer program that learns from the input people give the program in the form of, for example, the question to write an essay. MEP Kim van Sparrentak (GroenLinks) is not yet very concerned about its use among students. “While I was young, Wikipedia was just getting started, and I think we all used it avidly, and we didn’t come off very badly.”
Also read | ChatGPT hijacks education: how should schools deal with it?
AI is seen as a product and because there is a joint European market, there must therefore also be joint European rules, says Van Sparrentak. In addition, according to the politician, it is also important to arrange this jointly, because European countries are stronger together against the large companies such as Google and Microsoft that are involved in AI.
Also read | Microsoft invests billions in AI tool ChatGPT
According to Artificial Intelligence professor Stefan Leijnen, it is important that Europe legislates, but it is not enough, he says. ‘In Europe we are a little bit in the backseat and America is behind the wheel.’ It is therefore important that we develop high-quality AI products ourselves in Europe.
One way to tackle programs like ChatGPT is to have a transparency requirement. ‘If you call customer service, for example, you should always know whether you are talking to a human or an AI robot,’ says Leijnen. The current draft text of the AI Act now contains such a transparency requirement.
However, Van Sparrentak is concerned that ChatGPT’s underlying system will not fall under the law. It only looks at what AI applications are made for, not what it can potentially be used for. The MEP compares it to a piece of chalk. “A crayon is made for coloring, but we know that children will also put it in their mouth.”
Also listen | Looking ahead to 2023: this will make AI even smarter and tech regulations stricter
It is not entirely clear what ChatGPT was made for. ‘That would mean that the program will fall outside the law,’ thinks the GroenLinks politician. ‘These systems would then not have to meet standards so that they do not discriminate, for example.’
According to Leijnen, it is not always possible to incorporate standards and values into the software. Sometimes that is too difficult and it is therefore important that there is a certain social involvement in the development of AI.
ChatGPT came as no surprise to many scientists as they see a much broader trend. Leijnen expects the trends of personalization and generative AI to converge. ‘You will soon receive AI-made content that is also specifically aimed at you.’
‘This new form of personalized content will have an impact on privacy,’ Van Sparrentak imagines. A great deal of data is already being collected from consumers and, according to Van Sparrentak, we must keep a close eye on this to ensure that this does not get out of hand.
The merchant and the missionary
European legislation will also ensure that standards from Europe will also be used elsewhere in the world. ‘If you are an American producer of an AI algorithm; and you have to make AI that respects human rights for the European market, then you could also incorporate that directly into the American version,’ says Leijnen
It will take some time before all this legislation is actually in place. They are still negotiating very hard in the European Parliament. The law will probably be ready by the end of the year.