« first day (4757 days earlier)      last day (460 days later) » 

00:17
Also I did a bit of a survey, listening on YouGlish to AmE pronunciations of the phrase ride a bike. 20/30 had (to my ear) audible Canadian raising (i.e. the quality of the vowel in ride was different from the one in bike); only 10/30 didn't.
 
2 hours later…
02:10
Is it true that "melt" is usually intransitive, and would sound unnatural as a transitive verb?
I've just heard that the transitive verb for this context is "thaw" so
@DannyuNDos No. A snowman may melt, but he doesn't thaw. Frozen meat may thaw, but never melts.
Ice cubes melt. The ground (when frozen) will thaw.
Oh... That's the difference. Thx so much.
Metals may melt and become molten. Rocks too.
02:40
I disagree with Geoff Hinton regarding “glorified autocomplete” - by Andrew Gelman, professor of statistics and political science at Columbia University, 18 Nov 2023
Does ice thaw or melt?
> Be glad of water, but don’t forget
The lurking frost in the earth beneath
That will steal forth after the sun is set
And show on the water its crystal teeth.
32
Q: When should I use 'thaw' and 'melt'?

GenieThaw seems to be used when thawing something, like food. Melt seems to be used with ice. Some dictionary definitions seem to conflict with these statements. When should each word be used?

@alphabet I translated a lot of stuff about milk causing autism, but the jury is still out on that. Milk can promote the generation of anti-FRa antibodies, hampering folates from reaching the brain, because the animal FRa contained in the milk closely resembles the human FRa. The hypothesis is very interesting, because a sizeable proportion of kids with autism have been reported to have elevated anti-FRa antibody titres.
@CowperKettle I heard more about this. The success is that the body did not reject the other person's organ. There was no movement that the eye could do, and contrary to what most people would hope, the person could not get any sight out of the new eye.
@Mitch Yes, it's only the first step
Throwing in the stem cells was just a sort.of 'lets just do it and see'
@CowperKettle That is correct.
@CowperKettle yeah. It's totally amazing that it wasn't rejected. But the expectation by headline readers is that he can see now, which is not the case.
02:48
@alphabet For the "Milk as a factor in autism", see Ramaekers and Frye, they are researching the possible link between cerebral folate deficiency and autism.
However, one should keep in mind that there are many autisms, they are caused by different combinations of factors.
Also very unlikely... Stem cells haven't been shown to connect up new neurons for central nervous system nerves (like the optic nerve). They -have- shown some progress in peripheral nerves
> Recent research indicates that a number of children with autism generate folate receptor alpha autoantibodies (FRAA), which block transportation of folate across the blood-brain barrier. [...] Of 89 children, 30 (33.7%) were FRAA-positive. pubmed.ncbi.nlm.nih.gov/36112150
@CowperKettle That article in The Atlantic lists three meta-analyses which all found that the evidence for a link is incredibly weak.
@alphabet Yes, probably
Maybe it's a false lead, or maybe only a small % of people get their autism worsened by milk.
@CowperKettle Until there's a proper, large-scale study, I wouldn't pay it much attention.
02:55
However, one gene associated with cerebral folate deficiency participates in.. regulation of the immune response, and mice with a wrecked version of the gene start generating autoantibodies more readily. pubmed.ncbi.nlm.nih.gov/36112150
@alphabet There's a 2022 meta-analysis citing newer studies, but the meta-analysis was authored by the proponents of the CFD-autism hypothesis, so it's to be taken with a bit of salt. Or milk.
It's easy to cherry pick facts and come up with a plausible "mechanism" for any hypothesis you want to promote.
Yes, but at least it's no longer the "refrigerator mother" kind of hypotheses. It's something that could be tested and re-tested and so on, and refuted by tests.
Indeed.
My mom was a refrigerator. My dad was a microwave. My babysitter was a food processor.
I was raised by wolves. And household appliances.
@alphabet the Atlantic, New Yorker, and Harper's magazine are all very 'truthy' and 'sciency'. They all sound so believable. I never know whether to actually believe them or not
@alphabet wait so you're an orphan?
Your appliance parents gave you up for adoption and some wolves got you?
03:21
@Mitch It takes a village. One with both appliances and wolves.
@Mitch I don't trust 'em about science. But I trust the meta-analyses they link to, if they have them.
 
1 hour later…
04:37
I once came across a video with this Gabor Mate, and felt that he was fishy. The kind of psycho-celebrity who knows everything.
> Dr. Nikhil Chaudhary says that, for the vast majority of our species' evolutionary history, mothers probably had far more support than they currently do in Western countries such as the UK.
@CowperKettle Honestly, "ADHD is caused by childhood trauma" is nowhere near the list of worst ADHD hot takes
Imma go get myself un-traumatized just in case
> Mbendjele children benefit from the supplementary care of many people, but retain access to personal attention and consistency from a handful of key caregivers.
05:06
Sounds pretty good.
> Building capacity to produce ammunition in Europe has improved so significantly that “there might be parity” with American output by the end of next year if projections hold steady, said Camille Grand, who was NATO’s assistant secretary general for defense investment early in the war.
For the best, particularly given the possibility that Trump will be president again. Best not to rely on America too much.
Perhaps we should stop assuming that other NATO countries are just freeloaders.
Yeah.
We have kind of been freeloading.
It was assumed, after 1991, that Russia would never again be a threat, and so there was no reason to have strong armies, for noöne would ever attack us.
There's been a perception in American politics that certain EU countries have...rather little independent military strength and just expect America to fund their defense.
Which is probably true for some countries.
But apparently Germany is speeding up weapons production. Just don't call it "German rearmament."
@alphabet It used to be true if you assumed that Russia could attack us.
If not, then there was no need to have stronger armies.
European countries have been increasing their military budgets a lot over the past five years or so.
Which has only accelerated since 2022.
05:25
@Cerberus Perhaps Russia was never willing to attack because of NATO. There's a reason they won't invade, say, Poland, even now.
Granted, that poll is from 2014. But it doesn't look great for Europe (and it's only somewhat better for the US).
I do wonder, though, how those polls deal with translation. Presumably the precise terms in which the question is asked matter a great deal, and it's hard to say whether pollsters in different countries were really understood as asking the exact same question.
@alphabet Yes. Russia will not attack because of NATO. But that is exactly the freeloading: because America will come to out aid, Russia would never dream of attacking us. Why? Because America spends a ton of money on its army.
@Cerberus Yeah, that's what I meant also.
That said, Russia will still not attack us unprovoked even if Trump decided to leave NATO. It doesn't want a big war. However, it may consider itself provoked under certain circumstances, and attack. We cannot be sure.
@Cerberus Exactly. Putin can invent his own provocations.
05:40
Well, he doesn't want a big war.
He has attacked Ukraine because it was a justified war, and a short, victorious war.
"Justified."
I dunno. It's hard to say what Putin would do if NATO (say) suddenly ceased to exist. But he certainly would have been much more aggressive over the past couple decades (or however long he's been running things) without NATO.
@alphabet Ukraine was part of Russia; it was just that an illegal coup d'état, organised by evil Western powers, had seized power in the capital, and repressed the Russian culture and language of half of the Ukrainian population; most of them wanted to belong to Russia and would welcome the Russian army as liberators.
I think he mostly believed this.
@Cerberus There's also his fear of Ukraine joining NATO (which probably would have happened, eventually).
Yes, absolutely.
But that's just the reasons he says out loud. I dunno what's going on inside his head.
05:47
(Not sure whether that would have happened, though.)
@alphabet We have learned from e.g. Stalin's diaries that dictators often believe in the reasons they give in their speeches.
@Cerberus Stalin had diaries?
Perhaps not diaries, but documents found in his library with his notes.
Ah.
I just sort of imagine he makes decisions while walking around muttering to himself and laughing maniacally.
People like to believe in what they say.
Dictators more so than politicians in viciously polarised democracies.
 
1 hour later…
07:16
Word of the noon: chef de partie - line cook
 
1 hour later…
08:42
@CowperKettle Not to be confused with chef de parti: (political) party leader.
09:32
Colombia to Sterilize Pablo Escobar’s ‘Cocaine Hippos’. Apparently sterilizing hippos is harder than one might think.
10:01
[ SmokeDetector | MS ] Offensive body detected (47): Alternative verb to bullshitting‭ by White‭ on english.SE
10:24
:Oh dear, smoke detector is uoset.
10:37
Anyway Putin considers the demise of the USSR and the Waraw Pact the greatest geopolitical tragedy of the 20th century. So there are actions to fix—Chechenya, Syria, Georgia, Mondolva, Crimea (2914], 2022.
And let’s take out upstarts—Kodorovsky, Hong Kong, dissidents in general, Maiden Square, the writer in the elevator, the guy on the bridge.
They need out oil and gas more than we need them, because we have Chinese abd Indian markets.
Warsaw, that is.
 
1 hour later…
11:53
> Scientists have found that studying languages and other cultures can make a person more creative and compensate for a lack of communicative qualities. However, linguistic practices only enhance creative abilities in mentally stable individuals. tandfonline.com/doi/abs/10.1080/13670050.2023.2232521
12:13
@CowperKettle It started well but the punch line is disillusioning.
12:34
I've inhaled quite a bit of malathion and now I'm laughing for no reason
There was a spill
Not my fault this time I swear
@M.A.R. I'm very sorry to hear that! Take care!
Malathion is an organophosphate insecticide which acts as an acetylcholinesterase inhibitor. In the USSR, it was known as carbophos, in New Zealand and Australia as maldison and in South Africa as mercaptothion. == Pesticide use == Malathion is a pesticide that is widely used in agriculture, residential landscaping, public recreation areas, and in public health pest control programs such as mosquito eradication. In the US, it is the most commonly used organophosphate insecticide.A malathion mixture with corn syrup was used in the 1980s in Australia and California to combat the Mediterranean fruit...
It says that it's apparently not very toxic
You think they'd let undergrads near toxic stuff?
> In 1981, Malathion was sprayed over a 1,400 sq mi (3,600 km2) area to control an outbreak of Mediterranean fruit flies in California. In order to demonstrate the chemical's safety, B. T. Collins, director of the California Conservation Corps, publicly swallowed a mouthful of dilute malathion solution.[20]
@M.A.R. I thought they forced undergrads to work shifts in an underground uranium enrichment facility in Iran
It smelled like crap
I have a headache
@CowperKettle no that's only during lunch break. We're making rockets during class.
> Iranians enriching uranium
12:39
The guy in cyan is so relatable
@jlliagre I'm very mentally stable. I'm consistently yawning
@alphabet are you an ice cream or a burrito?
> What is the longest sentence in English?
- "I do"
Millennial humor
That was the favorite joke of Henry VIII
> Henry was best known for his six marriages
The guy changed wives faster than his underwear
"Bluebeard" (French: Barbe bleue, [baʁb(ə) blø]) is a French folktale, the most famous surviving version of which was written by Charles Perrault and first published by Barbin in Paris in 1697 in Histoires ou contes du temps passé. The tale tells the story of a wealthy man in the habit of murdering his wives and the attempts of the present one to avoid the fate of her predecessors. "The White Dove", "The Robber Bridegroom", and "Fitcher's Bird" (also called "Fowler's Fowl") are tales similar to "Bluebeard". The notoriety of the tale is such that Merriam-Webster gives the word "Bluebeard" th...
13:03
A wholesome story for everyone in the family
14:03
Apr 8, 2011 at 18:43, by Robusto
— Heh, here's another one for you then, also from Mencken:
"It is better to give than to receive — for example, wedding presents."
 
1 hour later…
15:25
> Our findings indicate that a shift from animal-based (e.g., red and processed meat, eggs, dairy, poultry, butter) to plant-based (e.g., nuts, legumes, whole grains, olive oil) foods is beneficially associated with cardiometabolic health and all-cause mortality. bmcmedicine.biomedcentral.com/articles/10.1186/….
@CowperKettle Everybody dies of something, though.
@CowperKettle that's just a label. You should absolutely not self-medicate though.
@Robusto Popeye died of iron toxicity
@M.A.R. See?
16:03
Weird Past-Tense Debacle: waked, woke, wakened, awakened, awoke, awaked. And I seldom feel completely comfortable with any of those.
Awokenated
Disagree.
@CowperKettle I doubt these numbers. They need to qualify them.
Jinx.
If you count manga, I'm almost certain Japan comes out on top.
@M.A.R. I don't have any doctor here who would investigate my symptoms
Considering that I was mis-diagnosed with diabetes for 23 years, I slightly doubt the Russian healthcare system. I have that fishy feeling about it.
16:16
Still better than India. Here you get medicine first then diagnosis 😂
@Vikas On the Nietzschean "What doesn't kill me makes me stronger" principle? ^_^
16:43
@Robusto Weird pronunciation debacle: duh BACK ull, or DEH buh cull
@Mitch The pronunciation of debacle is uncontested by people of letters.
Obviously, it's pronounced "french toast."
@Cerberus I'm impressed that GPT can produce a number that is sort of in range.
Part of theoretical computer science is trying to figure out what functions are computable with what mechanisms eg finite state automata vs turing machines (and all the variations in between). And NNs like GPT seem at a theoretical viewpoint to be very very complicated fintie state automata (unlike say a programming language like Python or C which are equivalent to Turing machines and can compute 'anything' (for a given definiiton of anything)
So when GPT shows that ... thinks up random calculation to do ...
16:59
@M.A.R. I am a Choco Taco
User: what is 497+203
ChatGPT: 497 + 203 equals 700.

User: what is 40097+20003
ChatGPT: 40,097 + 20,003 equals 60,100.

User: what is 4000000097+2000000003
ChatGPT: 4,000,000,097 + 2,000,000,003 equals 6,000,000,100.

User: what is 4,000,000,097+20,000,000,003
ChatGPT: 40,000,000,097 + 20,000,000,003 equals 60,000,000,100
It got the first three right. THat's amazing. You really can't expect a humongous FSA (finite state machine) to figure out (even statistically) the algorithm for addition (with carrying and such). But it really seems to act like it knows the algorithm.
The way ChatGPT's input data gets preprocessed makes it unable to read numbers digit-by-digit, making it very bad at arithmetic
But it doesn't ;know' the algorithm because that last example shows a very simple mod of the question makes GPT fail. But that it could fail so closely to the actual answer is a really amazing thing for an FSA to do.
@Robusto Hahaha it's funny because everyone hates their spouse but stays married anyway. (That seems to be the subtext of those jokes.)
@alphabet unable and bad are a bit extreme - it really does pretty good on the example I gave.
17:07
Hahaha marriage is a trap constructed to ensnare men in a lifetime of misery hahaha
Of course (the theoretical computer science 'of course), you can use an FSA to mimic what a Turing machine (TM) would do, up to a finite number of digits.
@Mitch I heard they pipe math questions to wolfram alpha to get a better answer. Apparently not in all cases…
@alphabet and the metaphor then follows that men would either gnaw off their own leg to get out of it... or lie down on the snow and let the cold drain them of life.
@Laurel That's one way to do things better, to combine systems. But I don't think that the the big OpenAI product of ChatGPT has the connection embedded officially.
I've heard that there are 'plugins' that you can add to chatGPT, and one of them is Wolfram Alpha.
The Wolfram Alpha interface itself is doing a lot of AI to help out (fixing typos in your math symbols, figuring out missing params, etc)
@Mitch IIRC the issue is that, before training data goes into ChatGPT, each word is replaced by a numerical token; the AI learns word-by-word, not character-by-character. Each number is treated as a single word, not as a series of digits, so it can't generalize about them very well.
But yeah, if chatGPT sees a mathy or sciency statement, it can rewrite it in a format that is official input to a more robust computer algebra system and solve it exactly.
17:19
It should be acceptable, when you hear one of those guys who likes to grouse about married life, to send them a list of local divorce lawyers so that they can have less to kvetch about
@alphabet It's not always whole words. A single big word is almost certainly several tokens. Big numbers are also broken into pieces. See platform.openai.com/tokenizer
@alphabet yeah the first step is to convert the string (the text input) into a sequence of numbers (makes it easier to process). And simple chat systems might only work on chunks that are space-delimited. I'm pretty sure Chat GPT has a slightly more sophisticated tokenizer that splits things a little more interestingly, maybe each digit is a token and words might get split into common pieces (like 'antidisestablishmenatraiansm' might get 6 or so chunks (even with my unintended misspelling)
@Laurel Yeah like that
@alphabet yeah the first step is to convert the string (the text input) into a sequence of numbers (makes it easier to process). And simple chat systems might only work on chunks that are space-delimited. I'm pretty sure Chat GPT has a slightly more sophisticated tokenizer that splits things a little more interestingly, maybe each digit is a token and words might get split into common pieces (like 'antidisestablishmenatraiansm' might get 6 or so chunks (even with my unintended misspelling)
@Laurel Yeah like that
Incidentally, kvetch is an interesting case of a loanword that doesn't fit ordinary phonotactic constraints and hasn't been nativized to do so.
hm... the interplanetary cyber network is having trouble sequencing packets as they fly back from the moons of Mars
@alphabet kvetch smvetch, try borshcht
As a native speaker, I find a lot of test questions like these difficult to figure out with all their negation switching:

"Do you not like to not squirm during a lecture/performance? Never, Some, not less than not much, Always"
17:46
@Mitch It's intentional, to avoid an issue called "response acquiescence bias."
i.e. different questions should go in opposite directions, since people have a bias towards answering "yes" to everything
18:06
@alphabet Sure but the cognitive load to extract what exactly you're signing up for is a lot.
18:17
Bing gets it all right.
But not the one about the Earth's circumference.
@alphabet Hey, don't shoot the messenger.
It really can't help itself.
This time, it gets it right.
It's so unpredictable.
18:39
@Cerberus Why are you torturing that poor ChatGPT?
19:30
0
Q: What resources do I need to consult before asking a question?

LaurelMany Stack Exchange sites require that users do some research before asking a question. Does English Language & Usage have such a policy? If yes, what resources should users consult before asking a question? Note: As suggested by Catija, this is a pared down list of resources that will be used i...

Also, does this lock prevent comments?
19:49
@Cerberus It did give the 'right' numerical answer (6371 km), even though it didn't follow your instructions about 'just the number'.
@Cerberus The interface doesn't give you the option of 'no randomness' (which it could)
@alphabet though I'm sure antibiotics would clear that right up
Yes!
Did you mean to lock it? I don't see any problem with allowing comments there.
20:28
@Mitch Ah well I wanted to avoid people posting new answers but I guess I'll just threaten to delete those instead :p
Wait, were you able to comment on the answer before I removed the lock?
@Mitch Exactly.
@Mitch Could it really?
20:47
@Laurel What's wrong with answers other than your own? I mean yours will probably be better but maybe there are some other quick refs that you just happened to leave out (I can't think of any).
I'm sort of considering giving a negative answer, eg 'don't use wikipedia or Urban Dictionary as authoritative resources'
@Laurel the screen shot shows that I just couldn't try...no UI available to me while locked.
@Cerberus I'm sort of ambivalent that it didn't 'follow directions'. I'm not sure how I feel about the pragmatics - are things said intended to be commands? Or is it OK to answer the question intended rther than the literal statement.
For example, maybe you were wrong in wanting just a number and no extra explanatory text 🤡
@Mitch Multiple answers could get overwhelming. It's supposed to be a short answer and it's already longer than I'd like
@Mitch I do have a note somewhere in the answer about ensuring that your sources are reputable
@Mitch That's a screenshot of the question
@Cerberus I'm fairly certain. Most language models have, for when 'inferring' (computing the next word or computing some following text) as opposed to training when it is ingesting the training text and slowly converging on all the parameters,.. most language models have a parameter called 'temperature' which is a level of randomness to add on a step. if temperature = 0, then it's entirely deterministic.
As temperature gets higher little changes start to come in on repeated runs of the same input. let's say for if the temperature is ~.3 you might get responses that are mostly semantically similar but use different words or slightly different syntax. For temp =1 you might get text that is perfectly grammatical and smeantically meanigful, but entirely irrelevant to what you just said.
@Laurel I get that it is desirable to have a single authoritative answer, but if you want people to not answer but instead edit community wiki style, then maybe you should make that obvious to everyone (either in a comment or in the question or answer itself.
@Laurel at the bottom of the screen shot is the place where one usually can 'add a comment' but it shows that that button isn't there, just the grayed out text saying 'no comments allowed/locked'
@Cerberus and since the programmatic function call with that temperature parameter is available, the UI could make it changeable.
I have a vague feeling you can change the temp parameter by some jailbreaking statements.
But to ensure you're testing correctly, you'd probably have to open a new ChatGPT conversation every time (usually the entire history so far in a convo is used for every new request)
21:06
@Vikas eh it's probably due to IITJEE alone.
For which self-reports are probably very, very off
@Mitch On SE, I would agree!
@Mitch Hmm depending on how the data are structured, that may make sense.
21:56
Wordle 884 4/6

⬛⬛🟨⬛⬛
🟩⬛⬛⬛⬛
🟩⬛⬛⬛🟨
🟩🟩🟩🟩🟩
Wordle 884 4/6

🟨🟨⬛⬛⬛
⬛⬛⬛⬛🟩
🟨🟨⬛⬛⬛
🟩🟩🟩🟩🟩
22:49
Wordle 884 4/6

⬜⬜⬜🟨⬜
⬜⬜⬜🟨🟨
⬜⬜🟨⬜🟩
🟩🟩🟩🟩🟩
23:01
#deluxewaffle78 2/5

🟩🟩🟩🟩🟩🟩🟩
🟩⭐🟩⬜🟩⬜🟩
🟩🟩🟩🟩🟩🟩🟩
🟩⬜🟩⬜🟩⬜🟩
🟩🟩🟩🟩🟩🟩🟩
🟩⬜🟩⬜🟩⭐🟩
🟩🟩🟩🟩🟩🟩🟩

wafflegame.net

« first day (4757 days earlier)      last day (460 days later) »