Opinion: AI Will Fail, and Resilience Requires Studying Everyday Mishaps

An opinion piece in Ingeniøren, written by researchers from the University of Copenhagen, DTU, AAU, CAISA and ROCKWOOL Fonden (among them Naja Holten Møller, Henriette Langstrup, Laust Hvas Mortensen, Nicki Skafte Detlefsen, Mogens Jin Pedersen, Thomas Moeslund and Andreas Bjerre-Nielsen), argues that AI will fail and that society cannot become resilient without studying everyday mishaps in technology use. The authors present it as their own view.

Their starting point is the CPR case, in which personal data on millions of people became accessible to unauthorised parties. They say it is being treated as a one-off, but that it illustrates a challenge that will grow as AI spreads. When public digital systems fail, the cause is rarely a single technical or human error, and the problem is usually much broader. As technology plays a larger role in society, they argue, frameworks must be created so that practitioners can learn from everyday errors and experiences.

Only the opening of the piece is available, so the specific proposals the authors go on to make are not known.

Two reader comments respond to it. One commenter says part of the current AI debate is shaped by a large influence campaign from the big AI providers, including myths and fear-mongering, and that easy access to the technology tempts inexperienced people into uncontrolled experiments. He agrees that an effective rules-and-control regime is needed and proposes that clients be able to take accountable responsibility for the consequences of their projects and that these stay within existing law rather than adapted law. The other welcomes the idea of investigating errors to avoid repeats, links it to a "crash commission" idea raised by an Ingeniøren/Version2 writer, and says the first step is getting the public sector to share information voluntarily.