Some basic terminology in robotics and AI
Bendel began by clarifying key terminology from his field and proposed some definitions, which he intended to be understood as “provisional results.” Robotics, he said, deals with the general concept, design, control, production, and operation of robots, for example as industrial and service robots; social robotics is concerned with the research and development of robots intended for interaction with humans and animals, although the two areas overlap to some extent.

Artificial intelligence, Bendel continued, is concerned with creating computer systems that can achieve, supplement, or surpass the results of cognitive processes in humans or animals. Generative AI refers to AI-based systems with which images, videos, audio material, and codes can for example be generated or simulations performed. These include text generators, which can use input or instructions (prompts) from the user to produce, summarize, evaluate, translate, edit, or paraphrase any kind of text. They can also be used as dialogue systems, such as chatbots or voice assistants. Text generators and chatbots such as ChatGPT or DeepSeek are based on large language models (LLMs) and other AI models. Multimodal LLMs can process and generate various formats.

Ethical issues concerning AI, robots, and machines
Technological ethics is concerned with moral issues arising from the use of technology, said Bendel. AI ethics in particular deals with the moral challenges involved in the development and use of AI systems. Robot ethics involves among other things the responsibility, rights, and obligations of robots. Bendel himself takes the view that robots, and computer and AI systems, are not capable of bearing responsibility. He is also a staunch opponent of rights for robots. Nevertheless, he noted: “I think it is necessary for robots to adhere to certain rules that we give them. These can also be moral rules.” Together with artificial intelligence and robotics, machine ethics is developing so-called moral machines. “These have no free will, no consciousness, no intuition, no empathy – they have none of these,” Bendel emphasized. “But machines can obey moral rules that we give them. That’s the crux of the matter.”

From the triad of the true, the good, and the beautiful to applications of generative AI
He then used the triad of the true, the good, and the beautiful, formulated by Plato, as a system for classifying artificiality by drawing parallels between the history of ideas relating to specific artificial creatures from antiquity and those of the present day, and by showing examples of generative AI. “My assertion is that the concept of robots – or of artificial creatures in general – was already familiar to humans several thousand years ago,” said Bendel.

Bendel made the leap from the ancient legend of Virgil’s brazen head to the chatbot ChatGPT, released by OpenAI at the end of 2022. “The public suddenly had access to a powerful AI tool and could generate texts and conduct complex discussions. But we were suddenly also dependent on it,” said Bendel. Chatbots have been in existence since the 1960s, he continued; a famous example was the ELIZA dialogue system developed by Joseph Weizenbaum. Users then experienced a veritable hype with chatbots and avatars around the turn of the millennium, when Artificial Life was a well-known provider of these digital communication partners and surrogates in the virtual world.

Based on his own research, Bendel demonstrated chatbots with which users could for example communicate in endangered languages such as Vallader, an idiom of Rhaeto-Romanic, and Basque (@llegra and kAIxo, 2023 and 2024 respectively). But because an increasing number of schoolchildren and university students are using text generators to write their assignments, an immense problem is emerging: “Pupils have ceased learning to write properly, and university students are no longer learning academic writing. They are also losing their individual voices.”

Hephaestus, the lame god of blacksmiths from Greek antiquity, created servant handmaids of gold to support him. Bendel drew a parallel between these “good” artificial creatures and examples of “good” chatbots. In a current project, for example, Bendel and his students are using multimodal language models for “animal whisperer” apps that analyze and evaluate the body language and behavior of animals in specific situations and then derive recommendations for their own behavior (e.g. when encountering cows while hiking). He critically remarked: “Generative AI can arrive at erroneous moral and other judgments and decisions, while at the same time bearing no responsibility for them.”

A further artificial creature created by Hephaestus was Pandora: although ignorant, malicious, and slothful, she was also beautiful in order to please Prometheus’ brother Epimetheus. In this connection, Bendel discussed the aesthetics of humanoid robots and referred to the “American smile,” which apparently dominates images of smiling people made with the image generator DALL-E 3. He also mentioned the fashion campaign launched in 2024 by the Mango company, which used digital models. “Generative AI produces beauty, but also stereotypes, and it gives rise to all kinds of distortions and prejudices,” Bendel criticized. “The laborious process of creation – learning from mistakes, improving over time, and the joy of finally seeing success: This all falls by the wayside.”

How high is the price of generative AI?
“As of 2022, generative AI became a tool for the masses; this led to a kind of democratization of AI. Opportunities opened up – also for the true, the good, and the beautiful,” Bendel summed up. “But at the same time, risks arose when generative AI fell into the wrong hands. Children, teenagers, and young adults are no longer learning basic cultural skills; criminals are stealing voices and identities and developing dangerous substances." Further problems relate to copyright and copyright protection, numerous opportunities for surveillance, and the power of corporations. And last but not least, attention must be drawn to the high level of consumption of electricity and resources needed for the training and use of generative AI systems.

“And what does this mean for our future? Should we abolish AI research? Should we take generative AI away from people again? Should we ban it altogether?” asked Bendel and added in conclusion: “This can hardly be the solution. But more than ever before, we should now work together to arrive at decisions regarding technologies that affect us all and create tools to prevent their misuse.”