A revolution in Internet research
The programme provides answers to almost every conceivable question – within seconds. It drafts poems and essays, writes programming codes or makes suggestions for marketing campaigns. And anyone so inclined can get it to provide tips for relationships. Since it first appeared in November 2022, the digital all-purpose weapon ChatGPT has taken our world by storm, changing everything with the aid of artificial intelligence (AI).
“For us as informatics specialists, ChatGPT has created a new reference system that is better than anything we had previously,” explains Prof. Benjamin Risse from the Institute of Geoinformatics at Münster University. As he sees it, the software is a new milestone in the development of AI models. ChatGPT is a text-based chatbot which interacts with humans and produces natural language. The abbreviation GPT stands for Generative Pretrained Transformer. In IT terminology, these three words stand for the way the developers designed the model: the text robot is composed of certain learning algorithms and was filled with enormous quantities of texts from the Internet. So far, the programme – which was developed by the American company OpenAI – relies on around 45 million pages of text. Using it via a chat mask is easy. All the experts agree: ChatGPT, and other systems like it, are not a new toy like a console – they represent a powerful new system. In its issue of March 4, the Hamburg news magazine Der Spiegel went so far as to describe it as a “new world power”. In the middle of March, OpenAI presented an expanded version of the AI language model which can also recognise and describe pictures.
These opportunities also present new challenges for research and teaching at universities. Is it allowed to use programmes such as ChatGPT when producing written work – whether it is an essay, a term paper or a doctoral dissertation? The answer given by the Rectorate of the University of Münster is unambiguous: “In cases in which students should write texts, it is absolutely forbidden to use text-generating AI systems. Any such use is an infringement of the law.” At the same time, the University management has set up a working group to monitor coming developments and draw up guidelines on how to use text-generating AI systems. The guidelines are to take account of both the risks and the opportunities presented by such systems.
“What we are now experiencing will change teaching and learning,” is the analysis of Prof. Stefan Klein from the Department of Information Systems at Münster University. “We are endeavouring to react quickly to this development,” he adds. In the summer semester of 2023 the Department is just one of several already offering courses on handling ChatGPT. “We need to raise students’ awareness and continue to promote critical, independent, responsible thinking,” says Klein. Prof. Ulrike Röttger from the University’s Department of Communication shares this view. “There have always been aids, the unfair use of which was forbidden. For this reason, we need to create an awareness of the problem on the basis of rules for good scientific practice,” she says. “It doesn’t accord with my view of humankind that students should indulge in widespread cheating. I much prefer to show them that scientific work is actually fun.”
Although ChatGPT supplies information from Google and Wikipedia within a matter of seconds, it still has weaknesses. Sometimes the software provides answers which are incorrect or not up to date. The reason is that it currently draws on data going up to September 2021. It does not, for example, know that Argentina won the World Cup in December 2022. In addition to incorrect material, the chatbot also supplies fictitious information, which experts term “hallucination”.
The fact that companies such as the software giant Microsoft are jumping on the bandwagon and investing billions in the OpenAI company goes to show the power which ChatGPT has in the sector. The No. 1 search engine Google also appears to be taking the development seriously. After ChatGPT appeared on the market, the New York Times reported that Google declared “Code Red”, a high alert level. “Searching the Internet will be revolutionised by ChatGPT as we will then be able to tap into content extremely quickly,” says Stefan Klein. “Looking ahead,” adds Benjamin Risse, “it will change the way we look for information.”
The geoinformatics specialist also expects AI technology to be used increasingly in many areas of life in the future. In his opinion, the field of so-called ‘computer vision’ – in other words, the AI technology which is able to process and understand images and videos – has enormous potential. “It will be of major importance in self-driving cars,” says Risse. Another area of AI is called ‘natural language processing’. “Basically, this is what ChatGPT or the translation software DeepL do,” he explains. Yet another area of AI is ‘decision making’ which, in Risse’s assessment, will be even more important. “There are already AI systems which help in making decisions in trading on stock-markets,” he explains. And something he considers to be just as important is ‘predictive analytics’, which uses data to forecast the probability of future events happening – something which could be of major interest to insurance companies.
The dynamics of the AI market are currently going global. This is being shown quite clearly, and not least of all, by the great interest being shown by the Chinese government in ChatGPT technology. In mid-March the Chinese rival to Google, Baidu, presented its own product to compete with ChatGPT: Ernie.
Author: Kathrin Nolte
This article first appeared in the University newspaper wissen|leben No. 2, 29. March 2023.