Mr. de Sombre, thank you for taking the time to speak with me briefly. You were particularly keen to oversee this study. Why do you think this topic is on so many people’s minds right now - including at Deutsche Telekom?
Telekom stands for digital communication, so the topic is a good fit for the commissioning organization. There is no doubt about its relevance. I only have to open a newspaper or look online to see its many different facets - from regulation in the form of the AI Act to the societal implications of using AI. I found the research questions very exciting. Especially because I now use chatbots regularly myself, both professionally and personally.
By the way: To what extent did you see yourself in the findings?
I could really relate. Like the people surveyed, I felt that fascination the first time I used it. The way the bot expresses itself alone - I was initially quite blown away by what AI can do. A few years ago, I would have considered that absolutely impossible. But after the WOW effect, I also experienced a sobering realization. In some respects, artificial intelligence is like a smooth talker that knows how to phrase things eloquently. Its manner of expression and choice of words can also inspire trust in users. But it is worth checking closely - because behind that polished surface, there is sometimes very little substance. Still, with healthy skepticism, my fascination continues to outweigh my concerns.
You are pointing to a finding that also emerged from the study. Two competing forces are at play: convenience and skepticism. When users are initially suspicious, they check the information generated again. But over time, the temptation to take the easy route could win out, and any information could be accepted at face value, right?
Yes, absolutely. I think it always depends greatly on how important it is for the information to be accurate. In a professional context, for example, the information generally needs to be correct. I can imagine that users check it more often there. In a personal context, people may accept information at face value more quickly and then pass it along. As the experts also say, that could pose a risk to democracy. Take opinion bubbles, for example: a highly selective slice of reality is repeated as fact.
Which findings surprised you most? Which ones less so?
I was not particularly surprised that younger participants and people with higher levels of education make greater use of generative AI. That was to be expected. What did surprise me, by contrast, was the rapid pace of change: In addition to the quarter of the population already using AI chatbots, another quarter can imagine doing so and will likely become users soon. And a large share of those who already use them expect to expand their use significantly. I was also surprised that broad sections of the population are calling for bans on AI applications in areas where I had expected strong support. For example, nearly half of users say it should not be used to diagnose diseases. This shows me that beneath the fascination, there is still a great deal of fear of the technology. Users remain very skeptical about anything that can have major personal consequences without being under their own control - diagnosing diseases, treating mental health issues, or selecting job applicants.
Finally, looking to the future: What is your forecast? More humans or more machines?
Definitely more machines. They will gain more ground in many areas of application and support us as assistants in many areas. There is a risk that certain cognitive abilities in society will be lost or weakened. But the experts in our study do not necessarily see this as a threat. If AI is used to outsource one’s own thinking, that is mental regression. But if I use AI with my own analytical judgment and knowledge and seek targeted support, then artificial intelligence can expand my own cognitive capabilities. Ultimately, it is a question of how it is used that determines whether it represents a step backward or progress. If I avoid thinking out of convenience: a step backward. If I use AI to support my own cognitive abilities: progress.
Interview by Kathrin Langkamp
