How can I competently work with generative AI?

In order to use generative AI in a competent and qualified manner, the user must be aware of the tools’ limitations and risks.

Like any other tool, generative AI also has limitations in its application, as well as certain weaknesses. In order to confidently and responsibly use of such tools in learning and teaching contexts, the user must be aware of these limitations and take them into account.

These issues, which are often complex and multi-layered, must therefore form an integral part of teaching AI-related skills, for example in the context of academic and scientific work.

One limitation of generative AI tools may include the timeliness of the data used to train language models: often, their training data only extend up to a specific point in time, meaning that more recent data and facts are not reflected in the system’s outputs.

Similarly, generative AI systems may reproduce (often unconsciously) prejudices and biases if these were present in the data on which the models were trained.

Further critical issues relate to the authorship of AI-generated texts and to questions of data protection and privacy, particularly with regard to the processing of personal data.

Overall, it should be noted that limitations vary across different AI systems, and therefore no universally applicable set of weaknesses can be defined.