Camembert or Pie Charts – Why AI Translation Fails
Camembert or pie charts? AI recognizes patterns but misses meaning. What seems funny on Duolingo becomes troubling when machines judge where humans should understand.
Category: Blog
A secret AI study. A biometric orb. A new internet ID.
After researchers sparked outrage with a secret AI experiment on Reddit, the platform considers adopting Sam Altman’s World ID system via a device called the Orb. But is biometric identity the solution, or just another dubious business model?
How AI Hijacks Human Connection
AI doesn’t just train on academic or artistic content. Increasingly, it feeds on blogs, guides, and independent journalism; any content that shows human care, credibility, or craft. Summarized and displayed in search results, this content becomes invisible at the source. Welcome to a world where creators are reduced to training fodder.
Why Human Learning Is Not Machine Learning
Machines improve performance. Humans seek meaning. This piece explores why learning is more than optimization – and what we lose when we confuse adaptation with transformation.
The Myth of AI Democratization
Some still believe that training large language models (LLMs) on copyrighted content is a form of “democratizing knowledge.” But when you look closely at how these models actually handle the material they ingest, the picture looks a lot less heroic – and a lot more extractive.
Premature compliance or belated authenticity?
Diversity once made it into glossy reports. Now it’s quietly shelved. Not because it must be, but because it’s easier that way. So what happens to values when they become inconvenient?
Klarna’s AI Whiplash: From Job Cuts to Human Epiphanies
From “AI can do all jobs” to “Humans are invaluable!”: Klarna’s AI journey is a masterclass in hype whiplash. But behind the cringe, the CEO’s rhetoric surfaces real ethical tensions. What happens when honesty about AI and jobs is no longer whispered in executive suites – but shouted?
Predicting Personality from Faces? Bad Science, Worse Ethics
Some say facial recognition and AI can assess your potential by analyzing your face. But what seems like innovation may be pseudoscience at scale – and a threat to privacy, fairness, and human rights.
Keeping AI Weird (for Safety Reasons)
AI makes mistakes differently from humans. And that’s a good thing. This post explores why we shouldn’t train machines to fail like humans and why weirdness might be an important safety feature of AI.
The data revolution devours their children
AI automates many things. And it is considered neutral. Will we finally achieve equal opportunities thanks to it? It’s not that simple: without human intervention, AI becomes a continuation of discrimination by other means. What’s more, LLMs are running out of food after years of data theft. They are increasingly feeding them with their own…
Rainwater Harvesting in Bangladesh
In 2017 I visited various NGO projects in Bangladesh and India in my role at the time as a board member of the Buchi Foundation. Upon return to Switzerland I wrote a piece on one of the projects together with Md Abu Hena Mostofa Kamal, who was field coordinator of HELVETAS at the time. The…
Getting to the root of food waste in bananas
The food waste of bananas created by consumers is only the tip of the iceberg. Even more waste is created at the farm level, where up to 40% of bananas are put to waste. The high percentage of waste at farm level puts an additional strain on the cost/income ratio of farmers.
