Stenobird

Episode

#350 How to Make Hard Choices in AI with Atay Kozlovski, Researcher at the University of Zurich

Podcast
DataFramed
Published
Mar 9, 2026
Duration seconds
4226
Processing state
processed
Canonical source
https://www.datacamp.com/podcast
Audio
https://dts.podtrac.com/redirect.mp3/cohst.app/pdcst/6G1A6D/episodes.captivate.fm/episode/0d3f89e9-4b0e-4671-819a-1a413f0951e5.mp3
JSON
/v1/public/podcasts/dataframed/episodes/350-how-to-make-hard-choices-in-ai-with-atay-kozlovski-researcher-at-the-university-of-zurich
Markdown
/podcast/dataframed/350-how-to-make-hard-choices-in-ai-with-atay-kozlovski-researcher-at-the-university-of-zurich.md

Actions

  • POST https://stenobird.com/v1/public/podcasts/dataframed/episodes/350-how-to-make-hard-choices-in-ai-with-atay-kozlovski-researcher-at-the-university-of-zurich/transcription-requests
    Idempotently request low-priority transcript generation for this episode.
  • GET https://stenobird.com/podcast/dataframed/350-how-to-make-hard-choices-in-ai-with-atay-kozlovski-researcher-at-the-university-of-zurich.md
    Read the agent-friendly Markdown representation of this episode resource.

Summary

As AI systems gain autonomy, the tension between efficiency and human control intensifies. This discussion explores the ethical frameworks needed to prevent systemic failures in high-stakes sectors like healthcare, welfare, and warfare.

Topics

  • AI Ethics
  • Meaningful Human Control
  • Algorithmic Bias
  • Deepfakes
  • Automation Bias
  • Digital Twins
  • Dual-use Technology
  • Normative Ethics

Highlights

  • Main idea: Responsibility requires a moral agent; AI systems cannot be held morally accountable for their actions
  • Failure mode: Automation bias and 'dual-use' technologies can lead to discriminatory policing and welfare fraud errors
  • Practical takeaway: Implementing 'meaningful human control' requires tracing decisions back to human-understandable reasons
  • Risk factor: The deployment of high-stakes tools without oversight disproportionately harms vulnerable populations
  • Critical skill: Maintaining epistemic diversity and seeking out opposing viewpoints is essential to avoid algorithmic bubbles

Chapters

  1. 1:00 The Paradox of Autonomy: Examining the contradiction between increasing AI autonomy and the necessity of human moral responsibility.
  2. 11:30 Systemic Failures in High-Stakes AI: Analyzing how errors in military and welfare automation lead to real-world harm and loss of human rights.
  3. 16:50 Tracing and Accountability: Discussing the technical and ethical requirements for ensuring AI decisions are traceable and justifiable.
  4. 27:20 Designing for Safety in Healthcare: A look at how intentional design constraints can prevent automation bias in clinical settings.
  5. 38:00 The Ethics of Digital Twins: Exploring the implications of using personalized data to create highly realistic human simulations.
  6. 1:04:40 The Future of AI Research: Reflecting on the tension between academic inquiry and the lucrative influence of big tech corporations.