Interview: Mete Ismayilzada - Unlocking AI's Creative Potential
Pushing the boundaries of innovation, Mete explores how artificial intelligence can transcend limitations to redefine creativity
Based in Switzerland at EPFL, one of Europe’s most vibrant and cosmopolitan science and technology institutions, Mete Ismayilzada is at the cutting edge of research into artificial intelligence. His work centres on understanding, evaluating, and enhancing the creative capabilities of AI systems, tackling complex questions about reasoning, abstraction, and innovation. Currently working on his PhD, part of a grander C-LING project funded by Swiss National Science Foundation and also a Research Assistant at the Idiap Research Institute, with an MSc in Data Science and a background in computer science, Mete bridges academia and industry, leveraging over five years of experience as a software/ML engineer and entrepreneur to bring real-world relevance to his research.
Mete's journey reflects a rare fusion of technical expertise and visionary thinking. From co-founding a tech startup to contributing to high-profile AI publications, his career illustrates a commitment to advancing AI's role as both a technological and creative force. In this interview, Mete shares his insights on the challenges and opportunities in AI creativity, the surprising discoveries shaping his research, and his vision for the future of human-AI collaboration.
Can you distill the key findings of your research on AI creativity?
Our review highlights the remarkable progress in the creative capabilities of AI systems over the past decade, driven largely by advances in generative AI. Modern AI models can craft jokes, write high-quality poems, generate long narratives, produce stunning images and videos, compose music, and tackle problems with creative solutions. The AI community has also developed numerous datasets, benchmarks, and metrics to assess these capabilities. Furthermore, we’re witnessing the rise of human-AI collaborations, where AI increasingly serves as a personal creative assistant, enhancing human innovation across various domains.
Despite these advancements, significant challenges remain. Current AI models, which are enormous and trained on vast amounts of data, often produce outputs that raise questions about the underlying processes—are these outputs the result of genuine creativity and robust generalization, or merely advanced interpolation and memorization? Studies reveal that AI struggles with real-world commonsense reasoning, compositional tasks, creative problem-solving, and abstract thinking. Models can copy large portions of training data, exhibit factual inconsistencies, generate less diverse content, and fall short of human experts in creative writing. Issues like copyright concerns and authorship questions further complicate the landscape. Our review also outlines future research directions aimed at addressing these limitations and fostering a deeper understanding and enhancement of AI creativity.
What has been the most surprising discovery in your research journey?
The most unexpected discovery in my work has been realizing how little we understand about the mechanisms behind the creative abilities of current AI systems, despite their impressive outputs. Much of the existing research focuses on evaluating the results these systems produce, whereas in fields like cognitive science and psychology, human creativity is often assessed not only by the outcomes but also by the processes leading to them. I was surprised to find that this process-oriented perspective is largely absent in AI research, though this is partly due to the black-box nature of these technologies. In our review, we advocate for a process-driven and multidimensional approach to evaluating AI creativity, incorporating aspects such as novelty, surprise, value, agency, and spontaneity. This shift could provide a deeper and more comprehensive understanding of AI's creative potential.
What directions will your future research take?
My future work will focus on evaluating and enhancing the creative capabilities of AI systems, particularly from a process-oriented perspective. I aim to uncover the fundamental limitations of current AI systems in the realm of creativity and explore ways to address them. Our review has highlighted significant challenges that remain unsolved, along with promising directions for future research, some of which I plan to pursue. As this is still an emerging field with vast potential, I am eager to contribute to its growth and excited about the opportunities it presents for deeper exploration and innovation.
Over the past year, how has your perspective on AI creativity evolved?
When I started my PhD last year, I held a somewhat pessimistic view of the research landscape, as the rapid advancements in AI seemed to leave little room for new contributions. However, over the past year, diving deep into the state of creativity in AI through extensive literature reviews and exploring hundreds of papers has transformed my perspective. This process helped me gain a clearer understanding of the significant progress made, the challenges that remain, and the exciting opportunities to make meaningful contributions as a researcher. My outlook has since shifted to one of optimism and enthusiasm about the potential for discovery and innovation that lies ahead.
Final remarks: Generative AI as a catalyst for amplifying human creativity
I believe generative AI acts as a catalyst for human creativity, amplifying our ability to imagine and create. By offering universal and democratized access, it empowers individuals from all walks of life to explore and enhance their creative ideas. Generative AI often serves as a powerful starting point, breaking through the initial barriers that can make creative tasks feel daunting. Once that first step is taken, people are more likely to find their flow and produce remarkable works. In this way, generative AI doesn't replace human creativity but enriches it, making the creative process more accessible and inspiring for everyone.