Back to All Events

Workshop on AI Vulnerabilities and their Societal Implications


  • Edinburgh Futures Institute 1 Lauriston Place Edinburgh, Scotland, EH3 9EN United Kingdom (map)

Workshop on AI Vulnerabilities and their Societal Implications

Join us in person on Thursday the 26th and Friday the 27th of September, at the Edinburgh Futures Institute, for a workshop organised by Dr Atoosa Kasirzadeh (Carnegie Mellon University) and Professor Des Higham (University of Edinburgh) in collaboration with the Centre for Technomoral Futures.

About the workshop:

We now live in a world where:

  • automated driving systems can misinterpret traffic "Stop" signs as speed limit signs when minimal graffiti is added,

  • artificial intelligence (AI) algorithms for medical imaging, microscopy and other inverse problems in the sciences can introduce hallucinations,

  • methods that aim to "explain" or "interpret" the decisions from an AI system are unreliable, and may themselves be attacked,

  • carefully tailored prompts can persuade chatboxes to generate offensive or sensitive material,

  • minimal edits to a deep learning classification tool or a large language model can hide undetectable surprises that change performance only on a specific, trigger input.

Within the world of algorithm development in AI, the past decade has seen a cat-and-mouse game between defence and attack, and it may be argued that the attackers currently have the upper hand. In a broader context this "Fallacy of AI Functionality" has raised issues around ethics, privacy and security; and it has clear relevance to the question of AI regulation.

This interdisciplinary workshop will address these issues by featuring accessible contributions from experts in algorithms that reveal or fix AI vulnerabilities, and researchers from philosophy, law and the social sciences with interests in the wider implications of AI shortcomings, with ample time for discussion and networking.

A key aim of the workshop is to bring together Edinburgh-based researchers with interests in AI vulnerabilities and their implications.

Speakers include:

The workshop will begin at 9.30am on September 26th and will finish at 12:30pm on September 27th. Lunch and a drinks reception will be provided on September 26th.

Click here to view the agenda

Click here to for speakers’ abstracts

Numbers are limited: applications to attend have now closed.

Important notice: If you have any questions regarding accessibility, please contact us at ctmf@ed.ac.uk

This event may be photographed/recorded, and images may be used for future marketing, promotional or archive purposes. If you would prefer not to be photographed, please let organisers know at the event.