German health insurance will reimburse a mental health app within days but cannot provide a therapist within six months. Last week, psychotherapy fees were cut by 4.5%. Baumol’s cost disease — originally about why string quartets get relatively more expensive — explains why the app gold rush and the collapse of mental health provision are the same phenomenon.
There is no seahorse emoji in Unicode. Ask a large language model to produce one and watch what happens. The failure is not a hallucination in the ordinary sense — the model knows what it wants to output but cannot output it. That distinction matters.
Anthropic has drawn a public line on military use of its models. The physics community spent the better part of the twentieth century working out what it means to draw that line after you have already built the thing. As a physicist watching this unfold, I find the parallels clarifying and the differences more unsettling than the parallels.
The intuition that feeding a language model more information improves its outputs is wrong often enough to matter. Here is why, and what to do about it.
AI detectors flag the US Constitution as machine-generated. They also flag technical papers, legal prose, and — with striking consistency — writing produced by autistic minds and physics-trained ones. The error is not in the measurement. It is in the baseline assumption: that systematic, precise writing is inhuman.
The universe has two measured expansion rates. One comes from the early universe, encoded in the cosmic microwave background. The other comes from measuring distances to nearby galaxies. They disagree by approximately 5σ — the threshold physicists call discovery. Every systematic error has been hunted down. JWST has confirmed the distance ladder. DESI has found hints that dark energy is not constant. Something is either wrong with our cosmological model or with one of two extremely well-tested measurement chains. We are not sure which.
A new video went viral last week: same question, “should I drive to the car wash?”, different wrong answer — the AI said to walk instead. This is neither the tokenisation failure from the strawberry post nor the grounding failure from the rainy-day post. It is a pragmatic inference failure: the model understood all the words and (probably) had the right world state, but assigned its response to the wrong interpretation of the question. A third and more subtle failure mode, with Grice as the theoretical handle.
I love concerts. I love setlists. I hate building the playlist manually afterward. But do I really? A small automation project, a Deftones show in Dortmund, and the question of whether you should automate something you kind of enjoy.
Microsoft confirmed this week that it hands BitLocker encryption keys to the FBI on receipt of a valid legal order. Windows 11 uploads them to your Microsoft account by default, without asking. For European universities that handle research data, student records, and HR information under GDPR, this is not an abstract concern. It is a structural problem. The answer is not a technical workaround. It is sovereign, publicly funded, openly licensed software — and a principle that the EU has articulated but not consistently practised: public money, public code.
A viral video this month showed an AI assistant confidently answering “should I go to the car wash today?” without knowing it was raining outside. The internet found it funny. The failure mode is real but distinct from the strawberry counting problem — this is not a representation issue, it is a grounding issue. The model understood the question perfectly. What it lacked was access to the state of the world the question was about.