RIP Tony Hoare, father of Quicksort. blog.computationalcomplexity.org/2026/03/tony...
12.03.2026 12:22
👍 40
🔁 15
💬 0
📌 0
RIP Tony Hoare, father of Quicksort. blog.computationalcomplexity.org/2026/03/tony...
Perhaps the first of an occasional series of persistently wrong ideas in Physics...
I was reminded yesterday of Maxwell's completely flawed derviation of his distribution, showing that it's more important to be original than right.
Can you see what's wrong?
www.cambridge.org/core/books/s...
When ChatGPT provides me with a link to a paper in a physics journal I can go and check it out. I’m wondering if you’ve used the latest models in earnest or if these are last year’s opinions…
#RIP Tony Leggett. He was a great physicist, winning the 2003
Nobel Prize in Physics. He contributed to several fields of physics, just look at his papers: journals.aps.org/search/resul...
He was also an outstanding & thoughtful referee for decades.
physicsworld.com/a/condensed-...
But increasingly LLMs ARE providing sources…
I'm being a bit more optimistic than that. When I'm using LLMs for physics exploration I ask for sources and then take a look at the sources if something seems promising. Often skeptics are asking "Are unicorns real?" and then complaining when the answer is not a straight "no"
Ok but THAT’S the problem surely? Not necessarily the LLM
Is measuring syndromes then using a decoder to infer the most likely error the best thing to do when trying to do quantum error correction? I prove cases where it is, where it isn't, and how to find the optimal threshold. Happy to get this out: let me know what you think! scirate.com/arxiv/2603.0...
Counter counterpoint: pluralism is *not* a lie
Brilliant article, simultaneously sobering and funny.
I enjoy solving LeetCode problems. Is this now as silly as enjoying long division? That makes me kind of sad. Perhaps the feeling of creative insight on finding a solution is no more than an artifact of my brain's rather fuzzy pattern matching
I think the Google scholar reader does a floating preview of a reference which I quite like. May be the only good thing about it. Not so keen on the AI summary of each section
Now the next question is whether the link accompanying a citation should go the reference section or straight to a doi
I don’t believe many people think of these as tools for telling right from wrong. Or at least I fucking hope not
Oh nice
I realise this is probably not a major use case but LaTeX or Typst to write mathematics and other technical content would be great...
Try going north
A small number of apps got to be big because they are very good and funnelling and retaining users. Blocking may not be the best way but I’m not sure why you think it would be totally ineffective
Ha! Do you like having to zoom in and scud about so you can read them on your phone?
PDFs are dead. HTML is the universal medium
Of course a human can also be inspired by ("pilfer") other work without being fully aware of the first place they saw a particular approach. I'm just a bit cautious of saying that this is qualitiatively different from what happens already. Evidently it is quantitatively different
Of course we DON'T actually cite everyone who contributed. There's a tacit statute of limitations on citations where the things deemed to be "common knowledge" are uncited. There's an interesting question of whether AI somehow expands that notion of common knoweldge. I'm not trying to be contrarian!
Also, how is the Born rule meant to come out of this? I assume you get something like a classical mixture of pointer states. But that somehow doesn’t seem satisfactory if you want to know more about where probabilities “come from”
Looks interesting. Definitely sounds less ontic than Everett. I guess I don’t get how an interpretation gets tested experimentally, especially when it’s based on ordinary QM of system plus environment
Surely the most boring belief a person can hold is that you are the lone voice speaking truth to power in your liberal milieu? Imagine him at a kid’s birthday sulking by the Wotsits
What is 2026's biggest quantum centenary? I'd vote for the statistical interpretation of the wavefunction, signalling definitive end of classical physics. Born won the 1954 Nobel prize for stating it most explicitly, though it was apparently no surprise in Copenhagen www.science.org/doi/10.1126/...
Nice! One distinction you might draw between conventional emergent thinking and ML driven methods is that the former pins its hopes on simplicity while the latter does not. We get taught that simplicity is good, because microscopic details are erased and universality emerges, but it’s not a given.
The fact that online harms are (of course) different in manifold ways from booze doesn’t mean that a similar two-pronged approach isn’t going to be necessary. That seems baked into the social contract of western societies
Hope this response is less annoying
Ok I was away for a bit but thanks for the response. I wasn’t trying to be smug. My point really is that there is a solid precedent for dealing separately with banning things for kids and getting adults to use them less. Risking oversimplification: feels a bit like whataboutism
I guess my usual response here is: are you ok banning alcohol and tobacco for teenagers?
