Aransentin
p ≥ 0.05 zombie
No bio...
User ID: 123
Aransentin
p ≥ 0.05 zombie
11mo ago
Huh! As an European I have always assumed it was the same all over the western world – a down-filled duvet in a removable bag akin to a big pillow case, and nothing else. Americans have multiple layers? Weird. Does the top layer not slide off or bunch together during the night?
Aransentin
p ≥ 0.05 zombie
11mo ago
You really need a bit of familiarity and prompt wrangling skill to get the most out of GPT. I've seen plenty of people testing it and fucking up basic stuff which makes them conclude that it's overhyped.
A common issue is not pressing the "GPT-4" button in the ChatGPT interface, and getting GPT-3.5 when you think you're running 4; and not being familiar enough with the interface to recognize that the blue-green icon means it's a lot dumber. E.g. Bryan Caplan fell for that.
A second issue is that GPT is at it's core a predictor, not a simulator. If you ask it something and it messes up the response, asking it something else in the same chat context means it will now start predicting that it's a character that makes mistakes. Resetting the chat fixes that. In the same vein, if you are giving it a quiz, having it predict a student response is suboptimal. Students make mistakes, you want it to predict the answer key.
Also, giving it few-shot examples improves it a great deal. If you have an example of a perfect query+response to a similar question, stuff it in the prompt itself! Then the LLM will know what kind of response you are looking for.
Aransentin
p ≥ 0.05 zombie
11mo ago
In the corner of the ChatGPT replies there's a little profile picture; if it's a blueish-green it's running a 3.X model; black means it's 4.
You might need to click on the image on the Caplan Twitter post to view the entire thing, it gets cut off in the preview window for me.
Aransentin
p ≥ 0.05 zombie
11mo ago
I am partial to the theory that advertising works by establishing a shared understanding of what consuming the product signals to others. It explains many puzzling behaviours, e.g. why Coca Cola would bother with ads – they want to steer how other people will percieve you when consuming their product, not necessarily create awareness or an immediate desire.
In that framework, the Mulvaney backlash makes perfect sense; the blog post I linked even talks about how brands generally don't have different messaging on different platforms since it would be directly counterproductive to the idea that you want to establish a common product image. InBev didn't understand it, tried different messaging for different groups, and is now suffering the consequences.
Edit: Oops, you said "apolitical" which this hardly is; at the very least not back when it happened!
I'd say that there's a somewhat reasonable chance that the Red Army Faction didn't commit suicide but where executed by the West German police, as the circumstances are pretty suspicious:
They managed to smuggle guns into their cells in one of the most secure prisons in the world at the time, in a wing built especially for them. The guards a few meters away didn't hear any gunshots, Baader fired thrice, missed the first two, and finally died from shooting himself in the neck from behind using his right hand, but he was left-handed.
There was no fingerprints found on Raspe's gun and no gunpowder residue, even though his gun was lying in his hand when he was found. Meinhof's hanging was suspicious as well (wiki article, too many things to list here). Möller stabbed herself four times in the chest, survived, and claims to this day that somebody tried to assassinate her.
Aransentin
p ≥ 0.05 zombie
11mo ago
I see Americans complaining online about how many spam telephone calls they are getting, to the point they don't even pick up the phone when the caller is unknown.
This is pretty alien to me; I live in Sweden and have literally never gotten a spam call in my life. (Maybe the reason for this is that the scammers naturally won't bother learning Swedish?)
So, potential silly lifehack: why not get a foreign telephone number from e.g. Denmark or something, and then never get spammed again? Presumably your calls will be more expensive as you'd be paying the international rate constantly, but this can be ameliorated by getting a plan where that's cheaper.
Aransentin
p ≥ 0.05 zombie
11mo ago
Before 2022 it was really easy, you could just buy a prepaid card with cash in a store and it'd have nothing tying it to yourself. Nowadays you need an ID. I checked a random telephone provider (Telenor) and they require you to use the national e-ID to buy a SIM card (or apparently upload a photo of your passport for international customers).
But yes, it's probably the limited market size and language barrier; especially since the prime target for scammers are old people who don't speak English very well.
Aransentin
p ≥ 0.05 zombie
11mo ago
Yeah, but presumably scammers don't care about that wherever they are.
Aransentin
p ≥ 0.05 zombie
10mo ago
Do roguelikes count? I've played a whole lot of them, like Dwarf Fortress, Cataclysm DDA, ADOM, and of course NetHack (for which I have to shamelessly brag that I hold the shared realtime speedrun world record in, from back when tool-assisted runs were allowed)
Aransentin
p ≥ 0.05 zombie
10mo ago
Thanks! If you're wondering why on earth we named it SWAGGINZZZ, it's due to us messing up the execution a bunch of times, and since we created a new account/player every time we pretty quickly exhausted all the good names we could think of. The final run was named that because we were just testing things and didn't expect that one to actually succeed..!
Aransentin
p ≥ 0.05 zombie
10mo ago
The SuperNova Early Warning System (SNEWS) is a network of neutrino detectors designed to give early warning of a near-earth supernova. They have a mailing list where you can sign up to be one of the first people to be notified when it happens. I signed up for it; just imagine the party trick you could pull off by getting that email one evening and pointing to Betelgeuse before it blows up!
Aransentin
p ≥ 0.05 zombie
9mo ago
Is anyone else interested in poetry here?
A thing I've noticed in poetry analysis that annoys me is what I've come to think of as "schizo" interpretations.
On the one hand, you have symbolism that was likely put into the text intentionally; e.g. in "Ozymandias" (which I assume you are familiar with) the famous "Look on my works, ye Mighty, and despair!" can be very reasonably interpreted as the onlooker ironically despairing that even the greatest ruler will eventually decline and be forgotten.
On the other hand you have stuff like "Scholars such as professors Nora Crook and Newman White have viewed the work as critical of Shelley's contemporaries George IV, with the statue's legs a coded reference to the then Prince Regent's gout". How reasonable is this interpretation? I think not very; if the poem had instead referred to the statue's arms or something, would there have been another possible tenuous interpretation to some other person or concept? Probably. You would need some sort of Bayesian intuition for this, as there is a "base rate" of possible random associations you could make – and for any connection less credible than that you're basically finding patterns in random noise.
It reminds me of how the famous schizophrenic programmer Terry Davis would "speak to God"; he had a random number generator that spat out words, and he'd do free associations between them. Textual criticism is rife with this. I suspect it's because there is really no incentive to find out the "truth" of a text, just finding cool associations that makes the reader look smart, and since there's no ground truth to verify anything it easily gets disassociated from reality.
To "fix" this, I propose a calibration game of sorts. One would write a text with actual symbolism and poetic devices, then publishing both the text and a canonical explanation for everything in it. Readers could then interpret it, and afterwards find out how much their interpretation missed the mark. If anyone wants to try this, I have done so with one of my old poems here.
(For the unfamiliar with meter, it's written in straightforward iambic tetrameter, i.e. each line consists of four pairs of unstressed-stressed syllables; with an ABAB rhyme scheme.)
Aransentin
p ≥ 0.05 zombie
9mo ago
In Sweden to pay our bills you can have a thing called "autogiro". This means you grant a company the right to pull a certain sum from your account each month, so you never have to bother with any administration for your bills. Pretty convenient, except I've been using it for all my monthly expenses so I've never had to log in to my bank at all.
Early this year was the first time I've logged in to my bank for years. I've had constant anxiety to see how much money I actually had, and finally I had to bite the bullet and actually look.
Turns out, I had MUCH more money than I expected. The relief I felt from that was pretty much indescribable; I've now proceeded to buy a condo much closer to the city centre, and have been going out to eat more or less every weekend.
Aransentin
p ≥ 0.05 zombie
8mo ago
Well, I'm of the opinion that the 'intended meaning' of a poem or piece of media is totally worthless and uninteresting
If that is so, you could very well just entertain yourself with a random word generator like Terry A. Davis, no?
The very best art taps into aesthetics or concepts that are to some extent transcendent
I don't necessarily disagree! Those concepts should in principle be discernible however, and above the "noise floor" of the random associations I'd label "schizo". I don't deny that there might be interesting unconscious features of works, just critiquing the tendency of critics to find signals in random noise.
Aransentin
p ≥ 0.05 zombie
8mo ago
No, but it was enough for the down payment of the loan.
Aransentin
p ≥ 0.05 zombie
8mo ago
In Demolition Man (1993), there's a scene where they mention Taco Bell. In the international version, this is dubbed over to Pizza Hut, presumably because it's more recognizable outside of the US. You can still kinda tell that the audio had been changed though, as their mouths don't match.
I'm pretty sure I wouldn't have spent much time reading the Taco Bell wikipedia page if that hadn't been done.
Aransentin
p ≥ 0.05 zombie
8mo ago
Yeah, there's a lot of text on the internet. With a pretty cursory Bayesian analysis, even with a 99.9% accuracy you're looking at a thousand false positives if you are combing through a million posts. Without some other thing to narrow it down, it seems reasonable that it'll not be possible from writing patterns alone.
Aransentin
p ≥ 0.05 zombie
8mo ago
You can use the "function calling" mode for that, using the API. It restricts the model to output JSON, so it doesn't get any opportunity to scold you about your questions.
I asked it to "Provide a list of (at least ten) races and their average IQ", and limited it to only return an array of objects with a "name" and "iq" field. The result was this.
Aransentin
p ≥ 0.05 zombie
8mo ago
Maybe the "real Londoners" refers to not using stock photography of posed models? The author here charitably had the title to work with and picked some stock image that looked noticeably inauthentic, and race never went into it at all.
Aransentin
p ≥ 0.05 zombie
8mo ago
As many said at the end of the deflated coup: "if it turns out that you can occupy one city, march in columns on Moscow, and then if you fail you will not suffer any consequences, then there may be many who will want to try to this themselves.
On the flip side, making a deal and then killing your competitor anyway ensures that people will be a lot less likely to make deals with you in the future, even in cases where you don't intend to defect. Having the killing be blatant then might actually work in your favour, as it won't be seen quite as sneaky and underhanded.
Aransentin
p ≥ 0.05 zombie
8mo ago
The biggest disappointment with the ending was
Aransentin
p ≥ 0.05 zombie
7mo ago
completely incomprehensible
I assume it's some sort of Straussian/obscurantist thing; intentionally hiding your points to prevent normies from being able to read it.
Much less charitably it's because his points are weak & often vibes-based, and any reasoning he makes would evaporate if stated explicitly and with any sort of rigour beyond Darkly Hinting to what you mean and letting the reader fill in the blanks. The few times he has written about things I am familiar with, the content really has been rather poor – take this article, for example, where he argues that a software that is only able to perform HTTP GET requests is safe, as such requests don't affect the server content. Anyone that has worked in web security know this is blatantly wrong, as there is probably hundreds of easily performed exploits and escapes for that weak of a sandbox.
Compensation in the US has more or less steadily grown since it started being measured in the 50s.
In pessimist/doomer spaces that want to make the economy seem worse than it is, e.g. Reddit, you frequently see charts that show otherwise. This is pretty much always due to dishonest stats, e.g:
-
Using "household income" instead of per-capita, which is confounded with shrinking household sizes.
-
Using inflators like CPI that doesn't take substitution effects into account (instead of e.g. PCE) and thus overstate inflation a lot if compared over a long period of time.
-
Not counting transfer payments.
-
Counting the decline in hours worked as lowered wages, and not as people choosing to work less when they don't need to.
-
Just completely making shit up, like this tweet that made the rounds a few days ago where real household income is compared to nominal rent prices.
Aransentin
p ≥ 0.05 zombie
7mo ago
"Real Hourly Compensation for All Workers" does not attempt to capture every possible thing in society that affects peoples finances, no.
(Though increases in cost of education will be reflected in the inflation, and as such adjusted for. Also the cost of the minimum viable laptop and smartphone required for getting a job is comparatively very low, and people get them anyway even if they weren't required – even the homeless have phones!)
Facetiously: Last Friday night. I am not a smart man when plastered.
Being a bit more serious, pretty much exactly 19 years old. I have a pretty extensive online footprint on various forums (civfanatics, criticalsecurity, various IRC servers... good times), so I can pretty accurately pinpoint when I stopped being super cringey and more or less myself.
(Tangentially around there I can also pinpoint when I discovered and consumed the Sequences, which purged me of a whole bunch of blatantly flawed thinking that I'd never have today)
More options
Context Copy link