@Aransentin's banner p

Aransentin

p ≥ 0.05 zombie

0 followers   follows 0 users  
joined 2022 September 04 19:44:29 UTC
Verified Email

				

User ID: 123

Aransentin

p ≥ 0.05 zombie

0 followers   follows 0 users   joined 2022 September 04 19:44:29 UTC

					

No bio...


					

User ID: 123

Verified Email

According to the Unicode CLDR (assuming the space character in the "localeKeyTypePattern" before the colon is the same as the space before the question mark, I'm too lazy to search around for the real spec location), Occitan and Breton does so too, somewhat unsurprisingly. The only other one that does is Adlam – a script used to write Fulani, a Senegambian language. Now that'd be a hard trivia question!

People have already noticed this IRL and people already accept it just fine, no radical anti-racist ideology needed. It's just the reality of the situation, sans any sort of ideology, that this sort of bias is fully and openly accepted.

Yes, but I could probably have been more clear: I am not claiming that society will demand AI models that necessarily treat men more fairly than we do today! A model with no anti-bias applied will consider men by by default to be extremely likely offenders, especially for violent crime. It is likely that any model can get a good training score by just looking at the gender and ethnicity, and if it's e.g. an Asian woman just let her off the hook immediately.

This effect will be sufficiently extreme to get noticed, and counteracted, by adding bias in favour of men or against those women – likely not enough to make the model as a whole to favour men more than women, but it will still be adjusted away from reality in a way that favours men! An AI that randomly decides to imprison men 50% of the time and women 10% of the time can still be biased against women if women commit 0.1% of the actual crime.

In sulla's initial reply he stated that the model will be biased in favour of blacks, and biased in favour of women, which are both true but only true if you use two different definitions: "manually adjusted to favour a group" or "returning different results for different groups, all else being equal". I assume people think my reply denied that women will be a favoured group under the second definition; I do not.

I don't think sulla is describing a crime model that's de-biased "naively," but rather one that's de-biased in the most likely way that it is to be de-biased, which is by explicitly putting the thumb on the scale

That's precisely what I meant with "naïvely", as opposed to other complicated schemes (such as the case with generative AI where you could potentially do tricks like adding "no discrimination" to the prompt or the like). Apologies if that was unclear.

I wrote a quick and shitty script to bruteforce English Sator squares from a big wordlist. The only one that makes any sort of limited sense is this:


t i m i d

i r a n i

m a d a m

i n a r i

d i m i t

I.e. "[the] timid Iranian madam that Inari dimit(=had dismissed)".

Edit: Of course somebody else had done the exact same thing and found the exact same square before. Oh well.

https://www.vatican.va has some deliciously shitty web design if you poke around a bit. The home page is pretty bad but not too crappy, but if you click around a bit you'll end up on pages such as this one, describing the pontificate of John Paul II. The way that they use alert boxes for information when you hover over the years are so utterly baffling that I'd immediately assume it was some sort of joke or shitpost if I saw it on any other website.

I find using GPT-3 as an "unblocker" works quite well. Insert the last few paragraphs you've written, and let it complete the text. The result isn't always very good, but you frequently get decent ideas on how to structure the next section.

Quite right, apologies for forgetting that. I haven't yet decided on whether to post about it whenever I've finished, but if so I'd probably mention the Baal thing without actually performing it. Another good reason is that it would also be the least kosher meal that'd nevertheless be permitted to eat if it would save one's life, and involving idolatry would mess that up.

Would you mind sharing what your idea is? I enjoy reading and thinking about software architectures.

Backslash, i.e: \*

One of the earliest programs I wrote was a random name generator for the ABC 800, a Swedish Z80 computer from the 80s; grossly obsolete by the time I got it but could run BASIC* well enough.

My current handle was generated from that, initially for some elf character I made for a MUD that i kept reusing everywhere.

* From what I can remember it was some cursed de-Americanized version of BASIC that e.g. modified the keywords and replaced the standard "$" string symbol with the international currency symbol "¤" to counter American cultural encroachment. As a fun aside, when the Swedish keyboard was standardized it got prominently featured as shift-4 due to this, though very few people since have ever used it.

I'm still wondering if there are exploits that use a (modern) browser alone, without relying on opening other software.

Yeah. Any time there's a zero-day exploit for a browser you can be sure that attacks will start using it fairly quickly. For example, cursorily searching online I found an example of two from last year targeting Chrome in the wild.

Edit: Here's another relevant article, from this year: "Google Patches Third Actively Exploited Chrome Zero-Day of 2022"

do browser plugins even exist anymore? I can't remember the last time I saw a page with a plug-in.

I'd say Flash and Java are completely dead for any moderately recent website, yeah. Still, computers might have the plugins installed; perhaps for some internal corporate website that will never been updated. Other than that, I'd guess the Adobe PDF plugin should be fairly common too.

The style that rdrama posts and then upvotes internally so it's visible may be distinct, but there's an obvious selection bias here in that the poster may very well just have been a low-quality rdrama user.

For all we know there could be a bunch of crap posts made by rdrama, like potentially this one, that just never rise to visibility there - resulting in a massively inflated view on what stuff actually gets produced.

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.

I'm fairly certain Scott does it because Yudkowsky did, and it spread to a bunch of other people on LessWrong. Yud himself has done it since forever, e.g. here ("the Other Reality [...]") in 1997. No clue where he got it from however.

The biggest disappointment with the ending was not having a scene where you give the photograph of the Insulindian Phasmid the cryptozoologists, proving that it is real.

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)

Yes, any sort of gatekeeping hurdle would improve it, though even mild inconveniences will significantly limit interaction from good users too. I wouldn't check e.g. this website very frequently if I had to spin up a VPN every time, for example.

It isn't really used for verification, no. Just a unique personal identifier for communicating with e.g. government agencies and banks; but there we're basically always using 2-factor authentication on top (a certificate you e.g. install on your phone, plus a code) so it doesn't really matter if my ID gets out.

Might this be testable?

IQ is valuable as it tracks actual intelligence fairly closely; if you are good at IQ-test-style puzzles you're very likely actually intelligent. This will be detectable in other things like complex creative tasks, and even practical areas like average income, crime rate, and so on. If the test was biased towards one gender by adding questions that the gender did better at, but weren't as strongly correlated with actual intelligence, you'd expect their IQ scores to be less correlated with the downstream effects of intelligence itself than it is for the other gender.

(Theoretical example: If men were a lot dumber than women in general, one could add "ability to lift weights" or the like to the IQ tests until both genders had the same average score. This would however result in IQ no longer being able to predict the ability of men to e.g. lead a company, like it still would for women.)

the baffling thing is, why is an openly trans officer willing to spy for Russia

Some sort of galaxy-brained double agent plot, perhaps? Leaking medical records of a few US officers is unlikely to matter too much for the war itself, so I'd guess it was mostly so they could be part of an exciting spy plot. If that's true, trying to set oneself up as a double agent is only marginally more delusional.

It seems like you’re asking the model to do two contradictory things, at the same time?

I mention that there are two approaches, targeting a politically correct world or the real one, not both at the same time!

I’m not sure what was wrong with the redcoats in Normandy

This was to contrast how ahistorical generic "knight" prompts are, akin to the AI unbiddenly putting redcoats in WWII if it was similarly bad at modern history.

A little project I want to do this year is try cooking an "Acherburger", that is, a meal that breaks the most kosher laws possible. There's some discussion about doing that online, but it's mostly low-effort stuff like "bacon-wrapped shrimp" etc. Lame! One can do so much more.

Here's the initial plan:

For the meat patty, combine as many treyf animals as I can. The supermarket in my city has a whole bunch, e.g. rabbit, kangaroo, alligator, even bear. I'd put a little of a bunch of them in, with the bulk being pork just so It doesn't taste too weird. Animal blood is forbidden as well; so I add a little bit of that too.

Onto this we'll add cheese, of course.

For frying, one can use suet instead of oil/butter, as that's forbidden.

Produce must be tithed before consumption, and you can't eat fruit during the first three years after planting. Outside of Israel, this isn't necessary unless you know for certain that it is the case; luckily I have a relative with a lime tree I know isn't that old, so I can add lime to the sauce and break that commandment.

Produce that may contain insects must be checked, or it is not kosher. Thus we don't do that for the lettuce, deliberately closing our eyes before putting some on the burger.

We'll eat it during the passover sabbath, so the fresh bread we'll buy is Chametz, Pat Akum, and Chadash (and of course the entire preparation of the meal is forbidden due to the sabbath). Naturally none of our utensils will be kashered either.

For the drink, we'll have wine. I have a bottle of Château Musar 2015. In addition to it being not kosher, 2015 was a Sabbath year in Israel, and since the wine is from Lebanon which counts as "Eretz Yisrael", it's not allowed. We'll also make it yayin nesech by pouring a little bit of it out in dedication to Baal. Before drinking it, I'll take a Nazirite vow to abstain from alcohol just to deliberately violate it.

The one rule I have some problem with breaking is Kil'ayim, that is "the planting of certain mixtures of seeds, grafting, the mixing of plants in vineyards [...]". This applies to Israeli produce only, and buying anything like that to make sure the seed were mixed during planting seems difficult. Sure, I can get spice mixes grown in Palestine in my local spice shop, but how can I be sure it actually broke any rule when it was grown?

Anything else I've forgotten about?

Much better than I anticipated!

I've figured that Lord Tennyson's "The Charge of the Light Brigade" could work as a Sabaton song, and it's indeed not bad.

(Also in Motown form, or 80's synth)

It it's intended to be a straightforward action/suspense movie, then it's not very good.

If it's intended to take the piss out of such movies, it's better. It's basically a John Wick type of story, but all elements being sufficiently shitty so as to just be above blatant parody. For example, there is a narration throughout the movie where the protagonist justifies his actions and puffs himself up with a sort of sophomoric "cool assassin guy" / "nothing personnel kid" persona (he literally calls civilians normies) as well making mildly unfunny quips. It's long unclear if it's just shitty writing; but in one scene his monologue is abruptly interrupted when somebody talks over him, so it becomes obvious that it's his actual internal monologue and just him being a blowhard. It just never actually blatantly winks at the audience, so you'll have to realize this yourself. (Another thing that points to this is that he constantly fucks up, especially when doing "cool ruthless assassin" things; like when he shoots nails in a guys chest during an interrogation and has an internal monologue calculating how slowly he'll die, but none of that works and he just dies instantly instead.)

Similar things are going on with the plot and character motivations; he wants to kill the people responsible for assaulting his wife (who is given the minimal amount of screen time to establish a motivation for the protagonist, and then is basically never thought of again). He leaves a pile of dead people in the wake, executing even innocent people without mercy to not leave evidence, but then in the end when he meets the big boss responsible for everything the guy gives him a piss-weak excuse of having no idea what is going on, and the protagonist just buys it instantly and lets him go.

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.

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.