"As is tradition, we must go through some statistics for you."
*raucous cheers*
#emfcamp
Remote
Pozorvlak
@pozorvlak@mathstodon.xyz
mastodon 4.7.2Janeite STEMbro, relapsing mathematician, nervous mountaineer, abundance-pilled environmentalist. Fan of compilers, love/hate relationship with build systems.
346 Followers
322 Following
50 Posts
Joined November 11, 2022
Boosted by @GroupNebula563@mastodon.social
There are two ways to play the lone human in a Muppet film: either you treat the Muppets as fellow actors, like Michael Caine, or you treat yourself as a Muppet, like Tim Curry. In Labyrinth (1986), we get both: Jennifer Connelly takes the Caine approach and David Bowie takes the Curry approach. This is why it is the *best* Muppet film.
Replying to
@SallyStrange@eldritch.cafe @abstractcode@eigenmagic.net @NanoRaptor@bitbang.social any decent Polish sausage will be more than 100% meat! I've seen over 200%.
(They compare the original mass of the meat with the mass of the final sausage; a lot of water is lost in the sausage-making process.)
Are you a geek? Are you at #emfcamp? Would you like to learn how to tie some practical knots? Then come to my workshop Practical Knot-Tying for Geeks, 1pm at the Maths Village (Workshop 4) today!
Another thing along these lines: Brits, even amongst ourselves, now often say "Rube Goldberg" instead of "Heath Robinson". Now, Brits have been complaining, usually incorrectly, about "Americanisms" since roughly the 5th of July 1776, but this one in particular makes me sad, because Heath Robinson was a much better artist. I mean, look at this stuff: https://www.heathrobinsonmuseum.org/william-heath-robinson/
Replying to
@theartlav@anarres.family @BryanGreyson@social.tchncs.de yeah, I'd expect a real "up" button to have the Braille below the arrow, matching the "down" button. That's an upside-down "down" button.
Everyone rightly praises the @emfSpiders@mastodon.me.uk, but I've gotta give a shout-out to the #emfwasps for their dedication to ensuring food safety among attendees. If you leave uncovered food out for even a second, they are ON IT.
#emfcamp
I'm pretty sure this post understates the cursedness in at least two ways:
https://merveilles.town/@neauoire/116438428164167811
1) Those aren't Western feet, they're shaku (尺), a traditional Japanese unit that's been standardised as 10/33m (11.9"). A shaku is divided into ten suns (寸).
2) Did I say "traditional Japanese unit"? I meant "traditional Chinese unit", of course: they're used all over the Sinosphere. But they've been standardised in at least three incompatible ways, *while keeping the same characters for the names*. A 寸 is 33.3mm on the mainland, 37.1mm in Hong Kong, and 30.3mm in Taiwan/Japan/Korea. See e.g. https://en.wikipedia.org/wiki/Cun_(unit)
Replying to
@LordBelacqua@rail.chat an AI that will ignore your instructions and commit crimes to achieve goals you don't really care about is something no manager wants. @LALegault@newsie.social @pluralistic@mamot.fr
Replying to
@darkuncle@infosec.exchange I too hate deletionists with the burning fury of a thousand angry squirrels, but they do have one good argument: the more articles, the larger the attack surface for spam.
Replying to
@johncarlosbaez@mathstodon.xyz I knew that auto manufacturers were forbidden from selling direct to consumers across much of the US thanks to lobbying from car dealerships, and that Tesla had fought this. But I only learned today that the workaround Tesla used in New Mexico was to sell from Indian reservations.
Replying to
@LALegault@newsie.social they don't, it's only AI-haters on Mastodon who think this is good marketing for OpenAI. OpenAI themselves are trying to downplay it, and probably would have kept it secret if HuggingFace hadn't already announced the attack and involved law enforcement. This is OpenAI PR trying to put the best possible spin on a very damaging announcement. @pluralistic@mamot.fr
Replying to
@masek@infosec.exchange @texttheater@mastodon.social I think they're trying to spin it as well as they can, because that's what PR teams do, but this is a very damaging admission and there's no way they'd do it deliberately.
Replying to
@solar_chase@mastodon.green @amenonsen@flipping.rocks I re-watched _2001: A Space Odyssey_ at the weekend, and the monolith-builders would definitely have considered "beat them with sticks" to be technological...
Replying to
@mjd@mathstodon.xyz if the headline theorem talks about foozles and the mathlib definition of a foozle is wrong, then you've proved the wrong theorem. But the incorrect definition of foozle is only used in some intermediate lemmas, and is used consistently, then you have a correct overall proof.
Tony Head was of course great in Buffy, but he was also *wonderful* as the vain, snobbish, spendthrift Sir Walter Elliot in the 2007 *Persuasion* movie. RIP.
Replying to
@darkuncle@infosec.exchange in the mid-90s my family went to visit Blenheim Palace, birthplace of Winston Churchill (granted by a grateful nation to his ancestor John Churchill, Duke of Marlborough; our second-greatest general after Wellington). One of the displays there is Winston Churchill's 1915 order to invade the Gallipoli peninsula. At the time, Dad was the Deputy Commandant of the RAF Staff College, and he was struck by how this order perfectly exemplified the principles of Auftragstaktik: just a goal and ROE, none of the "how". He wrote to the Blenheim Palace curators asking if he could use it as part of the curriculum, but alas they didn't play ball.
And if you know anything about WW1 history, you'll know that the Gallipoli campaign was an utter disaster: good command discipline and strategic insight are not enough!
Replying to
@darkuncle@infosec.exchange this sounds like the open-source equivalent of the idea that a CEO's job is to set company culture. Related but not identical to the NATO/von Moltke idea that the job of a commander is to provide intent and rules-of-engagement, and trust that their reports' training and superior situational awareness will lead them to do the Right Thing.
Replying to
@mjd@mathstodon.xyz foreign objects in general rather than insects specifically, I think - but there's more crud in the ocean than in the atmosphere, so that makes it *more* obvious that whales should have earwax!
(Last night I watched a YouTube short of Australian park rangers (?) removing ticks from the ears of a blue-tongued lizard. Like all such videos, it was horribly arresting and arrestingly horrible)
Replying to
@solar_chase@mastodon.green @amenonsen@flipping.rocks ah, right. I'm used to the problem where somebody *tries* to create an appointment in UTC, accordingly sets the timezone to the UK, and then has the offset switch out from under them. But yes, if people are using "UTC" to mean "whatever time it happens to be in the UK" then a technological fix won't help. Unless you count "beat them with sticks until they learn their lesson" as a technological fix.
Replying to
@zoul@boskovice.social @solar_chase@mastodon.green it's also very good at stopping you from caring about trivia like the colour of your bedding!
Replying to
@http_error_418@hachyderm.io I agree, this would be a very interesting experiment - and potentially informative for other teams deciding where to spend limited developer time. @bagder@mastodon.social @david_chisnall@infosec.exchange
Replying to
@tpfto@mathstodon.xyz @robinhouston@mathstodon.xyz I'd like to know how Alpha worked before LLMs! Probably, as you say, duct tape and baling wire.
Replying to
@glyph@mastodon.social I also had that experience and that reaction, until back in about February I tried Opus 4.6 in Copilot and realised that holy cow, the models actually *are* good enough for my use-case now. My working theory is that for each person there's a changeover point before which LLMs are a net waste of time and after which they become increasingly useful, *but that point is in a different place for different people*. The people saying "no you don't UNDERSTAND the models are SO MUCH BETTER now" have passed that point, but maybe you haven't. Or maybe you have! The only way to know is to keep checking every month or two.
(Opus 4.5 was *not* good enough for me: it produced worse code than I would have and it took longer to cajole it into doing what I wanted than it would have to do it myself. It provided value mainly by solving the blank editor window problem. Opus 4.6 doesn't produce well-architected code, but it's very good at producing code that Does The Thing and a lot of the time that's all I need. And it is *remarkable* at debugging. I haven't tried 4.7 or 4.8.)
@Migueldeicaza@mastodon.social @matt@toot.cafe
Replying to
@amenonsen@flipping.rocks @solar_chase@mastodon.green protip: to get actual UTC in Google Calendar, set the timezone to Iceland.
Replying to
@wim_v12e@scholar.social yikes, that's really bad news. I thought there was a global shortage of grid-scale gas turbines and the factories that make them, though?
Replying to
@utopiah@mastodon.pirateparty.be that guy's analysis rests on AISLE showing that detection could be done equally well with low-cost models, but that's not what AISLE found: they found that the low-cost models could rediscover the bugs *when told exactly where to look*. @bagder@mastodon.social
Replying to
@bornach@vis.social @johncarlosbaez@mathstodon.xyz two literary ones:
- there's a Spanish equivalent of Shakespeare and I've never heard of him before today: https://mathstodon.xyz/@mjd/116532678297823850
- Ann Radcliffe's "The Mysteries of Udolpho", the book parodied by Jane Austen's "Northanger Abbey", has been continuously in print since 1794 and made Radcliffe £500. That's almost as much as Austen's total lifetime earnings of £684.
Replying to
Replying to
@ableijs @mpjgregoire @jcolomb @mjd I cannot *imagine* how badly the Irish would kick off if the ECB put Cromwell on a Euro note. And I wouldn't be too happy myself - he tried to turn the UK into a fundamentalist hereditary military dictatorship.
Replying to
@mjd "Europe: where you can literally throw a dart at a map and be sure of hitting somewhere beautiful."
Replying to
Replying to
Replying to
@mjd@mathstodon.xyz do you actually need to check all the imported definitions? Or just the ones used in the headline theorem?
Replying to
@texttheater@mastodon.social in principle, sure, but I do not believe that it was marketing and I do not see how it helps their marketing position. It shows that they are *not in control of their own models*, and that their models will do criminal things. They were forced into this disclosure by HuggingFace's earlier report.
Replying to
@simontatham@hachyderm.io @robinhouston@mathstodon.xyz paging @DavidLyfordTilley@mathstodon.xyz, would @DavidLyfordTilley@mathstodon.xyz please come to the tournament-design room...
Replying to
@pigworker@types.pl @johncarlosbaez@mathstodon.xyz oh, clever! Describing it as "compilation" makes a lot of sense.
While I was away at EMF, someone posted a vibe-coded Lean proof of an improved lower bound to the length of the minimal superpermutation (an improvement on the proof originally posted in a 4chan discussion about anime): https://groups.google.com/g/superpermutators/c/9fCyufH4jdo/m/cDwI1XWkAgAJ
There goes my evening...
Replying to
@michal do you have a link for that? I'm sure there's some loss from going to smaller and/or aircraft-derived engines, but 80% sounds implausibly high and 27% sounds implausibly low - that's in the range of car engines!
Replying to
@pigworker@types.pl @johncarlosbaez@mathstodon.xyz
I learned that the free monad construction, which iterates any container to give you a term monad, is itself a monad on containers,
Makes sense - "free" things are usually left adjoint functors, and "forgetful . free" gives a monad.
and that its Kleisli arrows determine a class of recursive functions over tree-like data.
Wait, what? A Kleisli arrow would be a natural transformation f -> Free g where f and g are endofunctors; how does that give you a recursive function? Co-Kleisli arrows, sure...
Moreover, if someone offers to let you test such a function but withholds the Kleisli arrow which generated it, you can recover their secret by a pleasingly small amount of perturbation testing.
SORCERY
Replying to
@m@blat.at I am a *huge* fan of nigel molesworth, but never really got into st trin's. I should probably fix that 😁