“GenAI contributes to improving society” – Interview with Data Scientist and artist Mathias Riechert
Dr. Mathias Riechert is the founder of Mathias-Riechert-Art.com and serves as the Head of Data Foundation at Allianz P&C. He has spearheaded transformative initiatives at Allianz and BMW, catalyzing changes in data strategy, data architecture, data governance, and ethics. His art projects explore the interplay between AI and humanity. Mathias’s research interests revolve around AI adoption, AI ethics, transparency, and fostering transparent decision processes. At BARC’s DATA festival #Munich 2024 he will present to the audience his AI-generated images with a focus on societal challenges.
Mr. Riechert, at the DATA festival in Munich, you will present your project in which you show AI-generated images to raise awareness of societal challenges. Can you talk about your professional background/career and how it relates to the project?
Mathias Riechert (MR): As a former data scientist, my professional and personal background is directly intertwined with data. The idea to explore and present GenAI-generated images stems from my interest in how data affects our society and how new technologies like GenAI can utilize this data. I concern myself with the interface between human and system, specifically how media generated by GenAI is perceived by people and how this can generate motivation for using AI in socially and societally relevant areas.
Can you tell us more about the inspiration behind this project? What motivated you to start it?
MR: On a rainy afternoon in 2023, I wrote a book for my daughter using GenAI, in which she and her cello go on a personal journey. In addition, I generated images showing her on this journey. The resulting enthusiasm of my daughter made me realize the power of GenAI, especially through its easy accessibility. The project aims to use this easy accessibility as a basis for raising awareness of social and societal issues.
Can you briefly outline which tools and steps you used to create the images?
MR: It starts with researching the most important social and societal challenges for which AI could be a solution. For this, I had ChatGPT identify the top 10 challenges.
Based on this list of challenges, I had ChatGPT describe ten possible areas of application for each topic, so that a graphical form of it could be described. An example for Social Justice and Bias Reduction: “The image depicts a diverse group of individuals participating in a community engagement event facilitated by AI-powered platforms. These platforms provide opportunities for inclusive dialogue, collaboration, and civic participation, empowering marginalized communities to advocate for social justice and equity.”
I then used these 100 prompts in Midjourney for image creation, adding various styles, cameras, lenses, lighting, and perspectives. About 7000 images were generated in Midjourney.
After manual selection, image areas with errors in them were varied by regeneration in Midjourney and finally upscaled. The remaining 425 images were finalized with Photoshop. Finally, the images are displayed in a web interface to allow voting.
Were there challenges in implementing the project, and how did you overcome them?
MR: The manual review of the images is the most time-consuming process step. A possibility to automate Midjourney would be very time-saving. Furthermore, it would be elegant to be able to train one’s personal taste as a selection mechanism and apply it to Midjourney to speed up the image selection process.
How do you believe the perception and use of GenAI will evolve in the coming years, especially in relation to societal challenges?
MR: Technological development is rapid. When I started the project, public video generators were not yet feasibly usable. In less than a year, we see significant leaps in quality become reality, as exemplified by Sora. Media creation is just one facet. As it becomes so easy to generate content, I firmly believe that new solutions will be realized in social or societal contexts, much like ChatGPT “dreams up” today in its portrayals.
How do you define success for this project, and what long-term impacts do you hope for?
MR: The goal is to draw the AI community’s attention to socially relevant questions – through GenAI itself. It’s up to us how we use AI. This new efficiency can lay the foundation for using AI more where it’s most relevant to society.
Are there other projects or initiatives planned?
MR: Moving images have an even lower threshold for attention. Therefore, we would like to explore this further in a follow-up project.
Question: How do you evaluate the speed of developments in the field of generative AI in recent years? What milestones do you consider particularly significant?
MR: The pace of development of generative AI in recent years has been impressive. So fast, in fact, that one wonders during creation if an even more exciting technology will soon be available. Particularly significant milestones for me were the latest versions of Midjourney, as well as the current video generators Sora (OpenAI) and Emo by Alibaba.
What opportunities and challenges do you see in the increasing performance and availability of GenAI tools for creatives, researchers, and other areas?
MR: For traditional creatives, a significant shift in the business field is already visible. Humans are transitioning to a role as initiator, brainstorming partner, and reviewer/selector. This changes the level of creation significantly, but also allows for a much faster output when human and AI steps are combined. Instead of marking AI-generated media separately, in the future, due to the shifted volume ratios, marking of real photos is to be expected.
For science, I see enormous potential from GenAI. It can generate new types of data and content, enabling the addressing of previously effort-constrained unanswered questions. Experiments can be accelerated, which will enable more efficient research. Challenges are particularly around ethics, the increased need for quality control, and equality in data access.
How do you see the role of GenAI in addressing global societal challenges?
MR: GenAI can help with global challenges by generating innovative solutions in areas such as healthcare, environmental protection, and education. By analyzing data, it contributes to improving society by enabling more efficient processes and expanding access to resources. By allowing more people to participate more easily in finding solutions, we should be able to respond more flexibly to the rapidly changing environment.
What ethical considerations do you think are important as we use the power of AI to respond to these challenges?
MR:
- Transparency: It’s important that the functioning of AI systems is transparent, so users and those affected can understand how decisions are made and what their potential impacts might be.
- Fairness and Justice: AI systems should be fair and just to avoid discrimination and inequality. This requires careful design of algorithms and datasets to minimize biases.
- Privacy and Data Protection: Protecting privacy and secure data use are crucial to ensuring user trust and preventing abuse.
- Security and Reliability: AI systems should be safe and reliable to minimize potential risks to people and systems. This requires robust security measures and regular reviews of the systems.
Given the rapid development of AI, how can we as a society ensure that work with GenAI is conducted responsibly and ethically?
MR: We need to apply clear guidelines and standards for its use. It’s essential to set up education and training programs to create a broader understanding of the potentials, limits, and risks of GenAI.
Involving the public – as at the DATA festival – in discussions about the development and use of GenAI is crucial to consider various perspectives and opinions.
Thanks for the interview Mathias!
