Read the interview in German.

How do people gain an awareness or understanding of AI?

Matthias Spörrle: Humans define themselves in a strong way through intelligence, and rightly so: it is intelligence in particular that distinguishes us from other living beings and especially from other primates. Our intelligence has produced the special achievements that brought us to the moon and to the deepest part of the oceans. Because of our intelligence, we are the only species to have populated every climate zone on our planet - apart from the creatures we brought with us. We do not owe this to our strength or our senses, our intelligence is the central cause.

Starting with the Industrial Revolution, we have already given up the mechanical power of our hands and arms, for example, to machines. Today there are machines, a single one of which produces more power than all the humans estimated to have ever built the Cheops pyramid. So our mechanical achievements - which we owe to our intelligence - have long since far surpassed us in terms of mechanical power. Now this is beginning to happen again with our last core feature, our intelligence.

We humans notice what is special about this development, it goes to the heart of things for us, it is about the essential - in the literal sense. Accordingly, it is hardly surprising that we humans react so strongly to the topic of artificial intelligence. So our first awareness of Artificial Intelligence comes from an understanding of our own intelligence, we use our intelligence for guidance.

Isn't that misleading?

There are risks here, for example risks of anthropomorphization - in science we call it anthropomorphisation: we think that artificial intelligence is what we can do cognitively, but carried out by machines. But that is not yet the case. Artificial intelligences do not yet have a broader understanding of context; it is not an intelligence of the totality, but of individual performances. An example: If I give the image generation AI Midjourney the task of painting a father with three daughters and two sons in a landscape in a certain style, then I get an appealing image in this style of high quality, on which several people can be found. But the number of people does not correspond to the specification.

There are many people who can count to six and there are some people who could create a landscape in a desired style. But there are practically no people who have the ability to paint in that style, but at the same time do not have the ability to add small numbers. This is just an example to make clear: We humans take intelligence for granted and assume, for example, that if we have a high artistic ability, we have at least a rudimentary mathematical ability. This is true for human intelligence, but not for artificial intelligence. We do not yet have a general artificial intelligence, but we are dazzled by brilliant individual achievements. For most of us, our understanding of artificial intelligence is based on our concept of human intelligence, and that is wrong.

"The fact that people usurp power through tools is really nothing new. AI is also used in this sense. So far, the danger does not come from the technology itself, but from those who misuse it."

If AI is useful or comes across as fun, is acceptance greater?

Artificial intelligence is used in our everyday lives - whether we consciously perceive it as AI or not is an open question. But suggestions for product purchases, travel routes and recommendations for media we should consume are based on AI algorithms. And our own research as part of Computer&Automation AI RESEARCH was able to show that the more a person uses such technologies in everyday life, the more they experience themselves as taking advantage of the opportunities offered by the technology and are thus more likely to accept it. Interestingly, the people we interviewed also thought that they were better prepared for the disadvantages and risks of AI through their use. This can be critically questioned: By using AI, for example, I am not automatically better protected against AI-generated video forgeries.

Basically, a technology is accepted and used if it is experienced as pleasant, valuable, easily accessible, similar to the person using it and low-risk. Many of these characteristics are given for today's AI-based products or can at least be simulated well. The use of AI-based technologies will continue to increase and, according to the results of our research, this will further increase acceptance, a positive feedback loop.

Interview with Prof Dr Matthias Spörrle - Page 2

Creative Dyscalculia - Image generated by Midjourney based on the instruction: Imagine "A father with his three adult sons and two adult daughters on a boat on a lake in a beautiful landscape in the style of Caspar David Friedrich".

© Midjourney

What role does the social and professional environment play in terms of acceptance of AI? Can differences by industry or work environment be identified?

Of course, humans are also social animals in particular, we orient ourselves to others and look to the right and left before trying something; being laughed at in an unsuccessful attempt is less bad for some of us than for others, but it comes easily to no one. A society that is open to artificial intelligence, an organization that demands that its members engage with AI, colleagues who inspire me in their dealings with AI and a leader who sets an example for me in my work with the technology naturally increase the social tailwind so that I also engage with this technology.

There are big differences between the individual sectors, the technology-savvy ones are of course much further ahead here. But: we must not forget that intelligence is important in every area of our activities, as mentioned above. For example, archaeologists recently used AI to translate ancient writings from the kingdom of Akkad. Even if individual areas of our actions and work may not have had much to do with computers so far, they all have to do with intelligence performance. So there are varying degrees of readiness to use AI between different sectors and fields of activity, but I don't think that there are fundamentally fewer opportunities or less promising fields of application in the individual fields of activity. Intelligence is in demand almost everywhere. Those who do pioneering work here now and introduce the technology into new areas can position themselves successfully quickly and still with very little effort.

We are currently in the process of automating any work that is done on a screen. By tying the work to the screen, there is enough data available that we (still) need for the training of artificial intelligence. In concrete terms: Is your work predominantly carried out by means of digital channels or can it be carried out by means of these channels? Then sooner or later, data for training AI will be obtained from this. The activity will then first be supported by AI and then replaced. I think that, at least in economies like ours, the greater proportion of the total work done is already done digitally, i.e. on the screen - from programming, to virtual meetings and working with spreadsheet software, to operating the control room. All of these areas are thus accessible to the AI development that is currently underway. It will take a little longer for mechanical work, as less training data is digitally accessible here. The vast majority of us should be open to AI in our own field of activity and should actively engage with it.

"For most of us, our understanding of artificial intelligence is based on our human concept of intelligence, and that is wrong."

Dr. Matthias Spörrle is professor of business psychology with a special research focus on human decision-making processes in digital and economic contexts at the Private University Schloss Seeburg.

© Matthias Spörrle

In terms of AI, what developments and impacts can we expect in the future?

Let's venture a prediction: sometime between 2025 and 2030, the majority of the data - audio, video, text - on the internet that is currently still predominantly created and experienced by humans will no longer be generated by humans, but by artificial intelligences. We should not only think of texts from LLMs or images from text-to-image generators - Midjourney alone currently generates between 250 and 500 images per minute - but also of videos, pieces of music and voice messages created by AIs. Because a few sentences spoken by you are enough to mimic your speech; I can then use that to make you say anything, and then I turn that into podcasts, for example. This will have profound effects on several levels: The internet will largely become an environment whose sensory information no longer comes from us humans, but is only inspired by our training data. Thus, in one sense, AI will begin to train us. Do you still feel like you are in dialogue with me when you are no longer talking to me, but only to a chatbot - hopefully authorised by me - that speaks like me, pre-trained on a large speech corpus and re-trained with personal messages from me? How does your communication with me change when you know that you can no longer digitally distinguish between me and my bot?

Also, from then on, the majority of the information on which AIs are trained is generated by AIs. This could give rise to separate training paths for AIs, which over a longer period of time will detach themselves from the sensory inputs generated by humans. Currently, Model Autophagy Disorder (MAD), the increasing degradation of the results of generative AIs that are repeatedly trained with their own results, is being addressed in this context. The replacement of human-generated training data is already beginning, as individual applications work with artificial training data, for example for rare events with a consequently small database, or game-centered AIs no longer work with human games, but only learn from games against themselves. So here, humans are no longer even needed as a training data source.