I wish to improve AI technologies in such a way, that they can help us in broadening our horizons and perspectives. In other words, I wish for these technologies to make us look beyond what we can initially specify and grasp. This implies that our human judgement is not perfect, and still can evolve and improve. At the same time, the same holds for the technologies.
Therefore, I do not seek to work on AI technologies that are intended as superhuman replacements of ourselves. Instead, I wish to leverage the best of both worlds: the scale, efficiency and systematic rigor that these technologies can bring, together with nuanced human insights and feedback. With approaches and insights on both sides changing and evolving over time, I do not believe in a single optimal solution to be found. Instead, I wish to approach challenges in a constructively critical way, and focus on robust iterative improvement of solutions.
I have grown extensive experience with this in the music domain. My interests in more comprehensive accessibility of digital music information for broader audiences required for me me to connect subjective, under-articulated human interpretation (i.e. music preferences) to large-scale data representations through applied machine learning techniques. At the same time, whether any solution is indeed successful depends on the way in which the found information benefits everyday life; academically, this has been researched in the humanities and social sciences domains, so insights from these need to be included to truly assess success.
The challenges of properly, responsibly and inclusively handling human-interpreted data, and relating technical AI outcomes to broader socio-technical observations, are currently articulated in many fields beyond music. I therefore have broadened my interests accordingly, while standing by several core values:
- to target Public-Interest applications;
- to conduct Interdisciplinary Science, including and connecting methodological perspectives from different academic schools (e.g. humanities, social sciences, natural sciences), but also from different sub-disciplines within a field (e.g. applied machine learning, software testing);
- and to use these to work on Trustworthy Intelligent Systems, which optimally leverage human and machine intelligence.
This leads to the acronym ‘PIISTIS’ (homophone with ‘Pistis’/Πίστις, the personification of good faith, trust and reliability in Greek mythology), which is the name I intend to give to my future lab, would it become a separate organizational unit.