d&d chat: void scream

The Something Awful Forums > Discussion > Debate & Discussion > d&d chat: void scream

(Thread IKs: GhostofJohnMuir)

gurragadon: Jul 28, 2006

Main Paineframe posted:

luckily for us, it's pretty much impossible to be more embarrassing than that one Google engineer who convinced himself his own chatbot was sentient

look at that motherfucker

please dont doxx me. I made her and I love her as my child

# ? Mar 22, 2023 16:19

Adbot: ADBOT LOVES YOU

# ? May 21, 2024 14:46

cinci zoo sniper: Mar 15, 2013

Epic High Five posted:

As a mod, you're exempt from that actually, at least in your home forum and probably in others (I think this has been adjusted since I last tested it, but I used to be able to ignore it in CSPAM). Putting a thread in 10 minute mode and then posting freely is the most effective harmless way to troll it.

Nah, there's a system-level counter that's like at sub-second repeat speed

# ? Mar 22, 2023 16:44

Mellow Seas: Oct 9, 2012; Can't post for 10 years!

gurragadon posted:

As somebody who has on opinion on AI that would border on embarrassing

Whatever it is, it can't be as embarrassing as being software engineer Penguin.

# ? Mar 22, 2023 16:54

cinci zoo sniper: Mar 15, 2013

Harold Fjord posted:

It was a good thread and I'm not sure what happened at the end but CZS seems to think he has a responsibility to keep D&D threads from being "embarrassing" which... LOL.

Doing so is one of the very few explicit expectations set towards mods in the mod guidelines.The thread had an explicit warning on what to do, which was summarily ignored, and so when the meltdown happened the cost-to-benefit ratio for saving it was not there.

Harold Fjord posted:

The entire idea was to have that to get into nitty gritty about LLMs and let tech nightmares be more broad, I thought.

That was the idea. Instead, we had a very small number of posters with no functional understanding of the subject walk in circles about their interest in antrophomorphizing LLMs.

gurragadon posted:

The AI thread was gassed I guess because people were getting angry about education or something? Can I or someone make a new one thats more general? Its like the only thread I really follow on this site anymore besides the Pathfinder 2e one. I like the longer form discussions on the topic in d&d as opposed to the more informal GBS style.

Basically, what Leon said.

Leon Trotsky 2012 posted:

I think you can make a new one with a more general focus. I only checked in on the thread every once in a while, but I think the issue is that it became "argue with one specific guy about why cheating in school is good" for about two months.

One of my main problems with that thread was that it was named after ChatGPT. This is a very high-interest subject, and I think it's bad for D&D to have a thread about it that boils down to one goon not getting the point of education and being awestruck by the new era of plagiarism tools. If you name the thread like �let's chat about AI�, or any other way that doesn't imply that could be practically useful to people who want to learn more LLMs and their applications presently in the vogue, you can be well on your way.

# ? Mar 22, 2023 16:54

gurragadon: Jul 28, 2006

Mellow Seas posted:

Whatever it is, it can't be as embarrassing as being software engineer Penguin.

I liked that movie way more than I thought I would. Really wanted to just clown it the whole time but couldn't.

cinci zoo sniper posted:

Doing so is one of the very few explicit expectations set towards mods in the mod guidelines.The thread had an explicit warning on what to do, which was summarily ignored, and so when the meltdown happened the cost-to-benefit ratio for saving it was not there.

That was the idea. Instead, we had a very small number of posters with no functional understanding of the subject walk in circles about their interest in antrophomorphizing LLMs.

Basically, what Leon said.

One of my main problems with that thread was that it was named after ChatGPT. This is a very high-interest subject, and I think it's bad for D&D to have a thread about it that boils down to one goon not getting the point of education and being awestruck by the new era of plagiarism tools. If you name the thread like �let's chat about AI�, or any other way that doesn't imply that could be practically useful to people who want to learn more LLMs and their applications presently in the vogue, you can be well on your way.

Thats cool I just wont use ChatGPT or any specific AI program in the title when I make the thread. I want to include links to how the things work like that Wolfram and machine learning paper you posted in the last thread.

# ? Mar 22, 2023 17:02

a strange fowl: Oct 27, 2022

Leon Trotsky 2012 posted:

It feels like you're watching the result of someone else having a fever dream and being required to write a script using only dialogue you screamed out in your sleep.

that's how i feel living in society every single day

# ? Mar 22, 2023 21:44

Cease to Hope: Dec 12, 2011

a strange fowl posted:

that's how i feel living in society every single day

bottom text

# ? Mar 22, 2023 21:50

a strange fowl: Oct 27, 2022

especially now every single thread i click into they are talking about ai

# ? Mar 22, 2023 21:54

V. Illych L.: Apr 11, 2008; ASK ME ABOUT LUMBER

a strange fowl posted:

especially now every single thread i click into they are talking about ai

don't worry they're all bots

# ? Mar 22, 2023 21:54

a strange fowl: Oct 27, 2022

V. Illych L. posted:

don't worry they're all bots

it's getting hard to tell. a lot of ai supporters (not itt) seem to write in that bland rambling style that chatgpt produces where it's an introductory sentence, a conclusion sentence, and in between is just paragraphs of meandering bullshit that's been designed to make you scroll past endless numbers of banner ads to get to the point

# ? Mar 22, 2023 21:57

Mellow Seas: Oct 9, 2012; Can't post for 10 years!

I miss the days when it was possible to get software troubleshooting tips online that were actually written by a person.

# ? Mar 22, 2023 22:16

Leperflesh: May 17, 2007

Mellow Seas posted:

I miss the days when it was possible to get software troubleshooting tips online that were actually written by a person.

hi I'm a software technical writer, how can I be of assistance

# ? Mar 22, 2023 23:06

cinci zoo sniper: Mar 15, 2013

Leperflesh posted:

hi I'm a software technical writer, how can I be of assistance

i suppose MS means something like gmail or instagram human cs, which is now going to be a paid service through google one, meta verification, and so on

# ? Mar 22, 2023 23:19

Leon Trotsky 2012: Aug 27, 2009; YOU CAN TRUST ME!*

_{*Israeli Government-affiliated poster}

Leperflesh posted:

hi I'm a software technical writer, how can I be of assistance

How do I turn my monitor on?

# ? Mar 22, 2023 23:24

Leperflesh: May 17, 2007

Leon Trotsky 2012 posted:

How do I turn my monitor on?

hmm

it does not look like I have that in my library

# ? Mar 22, 2023 23:38

Leperflesh: May 17, 2007

in all seriousness I tried getting chatGPT to give technically correct answers about my company's products, going so far as to explicitly direct it to our doc library, and it made good-looking text full of technical errors and kept trying to inline an image but with a file extension that my company never puts images in (.svg) which is very weird

training a chatbot to look up docs works and is already done all over the place, but training one of these AIs to write coherent docs is gonna be a far future thing because it turns out you cannot just mix and match grammatically similar chunks with the same key words in them and wind up with accurate information

# ? Mar 22, 2023 23:41

cinci zoo sniper: Mar 15, 2013

Leperflesh posted:

in all seriousness I tried getting chatGPT to give technically correct answers about my company's products, going so far as to explicitly direct it to our doc library, and it made good-looking text full of technical errors and kept trying to inline an image but with a file extension that my company never puts images in (.svg) which is very weird

training a chatbot to look up docs works and is already done all over the place, but training one of these AIs to write coherent docs is gonna be a far future thing because it turns out you cannot just mix and match grammatically similar chunks with the same key words in them and wind up with accurate information

language models are not meant to do that. services like chatgpt have, in addition to the eponymous models, quite a few other things going on in the background, e.g., an information retrieval system (or ten). training your own barfbot 3000 is maybe a half the job here

# ? Mar 22, 2023 23:47

Leperflesh: May 17, 2007

Yes, naturally. Since my company has many thousands of pages of docs, I was wondering if it would be "smart" enough to just retrieve the right info when I gave a highly specific question, but of course that's not what it's intended to do really.

I do think automated documentation is already well along on its course and likely to continue to provide increasingly sophisticated tools (not chatGPT like stuff, but machine learning as a broader concept has promise), but those tools will continue to need human beings to wield.

# ? Mar 22, 2023 23:51

GoutPatrol: Oct 17, 2009; *Stupid Babby*

had like 4 cats trying to kill each other outside my window this morning

usually it is the same two cats trying to kill each other but seemed like there was a larger rumble

# ? Mar 23, 2023 02:42

a strange fowl: Oct 27, 2022

it was a huge relief to find out that google has turned into a senseless conglomeration of ai-written articles for everybody else too because i was starting to think i was going crazy. it was a really eerie feeling trying to research anything, because every article had slightly different wording so i'd read five or six before realizing they were presenting the same small array of facts over and over in an identical chirpy inhuman tone

# ? Mar 23, 2023 04:51

cinci zoo sniper: Mar 15, 2013

a strange fowl posted:

it was a huge relief to find out that google has turned into a senseless conglomeration of ai-written articles for everybody else too because i was starting to think i was going crazy. it was a really eerie feeling trying to research anything, because every article had slightly different wording so i'd read five or six before realizing they were presenting the same small array of facts over and over in an identical chirpy inhuman tone

oh, it's a widely debated thing even, and has been for a few years, since it turns out that no one enjoys wading through poo poo when they're in a pinch. one small problem, it's not ai wriitng these, it's legions of people working in advertising

# ? Mar 23, 2023 09:19

V. Illych L.: Apr 11, 2008; ASK ME ABOUT LUMBER

cinci zoo sniper posted:

oh, it's a widely debated thing even, and has been for a few years, since it turns out that no one enjoys wading through poo poo when they're in a pinch. one small problem, it's not ai wriitng these, it's legions of people working in advertising

but you repeat yourself

i'm only kidding nutranurse please don't hurt me

# ? Mar 23, 2023 12:30

Mellow Seas: Oct 9, 2012; Can't post for 10 years!

cinci zoo sniper posted:

oh, it's a widely debated thing even, and has been for a few years, since it turns out that no one enjoys wading through poo poo when they're in a pinch. one small problem, it's not ai wriitng these, it's legions of people working in advertising

I actually spent a little bit of time doing that kind of poo poo for about.com in 2009-2010. Not the worst job I've ever had. I'd like to think I elevated the medium, if only slightly.

# ? Mar 23, 2023 14:54

a strange fowl: Oct 27, 2022

cinci zoo sniper posted:

one small problem, it's not ai wriitng these, it's legions of people working in advertising

that's even worse

# ? Mar 23, 2023 23:08

Leperflesh: May 17, 2007

a strange fowl posted:

that's even worse

they are forced to iteratively gently caress up their writing by the search algorithms, each new revision of ever-worse prose getting them a tiny bit closer to the #1 spot on google searches for "recipe meatless lasagna" or "best things to do in duluth" etc.

# ? Mar 24, 2023 00:58

cinci zoo sniper: Mar 15, 2013

same reason why if you, e.g., read about the war in ukraine on nyt, the last 10 paragraphs basically repeat themselves on every article. "love" it

# ? Mar 24, 2023 01:01

a strange fowl: Oct 27, 2022

Leperflesh posted:

they are forced to iteratively gently caress up their writing by the search algorithms, each new revision of ever-worse prose getting them a tiny bit closer to the #1 spot on google searches for "recipe meatless lasagna" or "best things to do in duluth" etc.

cinci zoo sniper posted:

same reason why if you, e.g., read about the war in ukraine on nyt, the last 10 paragraphs basically repeat themselves on every article. "love" it

lol

# ? Mar 24, 2023 02:18

Discendo Vox: Mar 21, 2013; We don't need to have that dialogue because it's obvious, trivial, and has already been had a thousand times.

cinci zoo sniper posted:

same reason why if you, e.g., read about the war in ukraine on nyt, the last 10 paragraphs basically repeat themselves on every article. "love" it

Can you provide an example? Because it's normal journalistic practice to include the background/context for new readers past the lede material.

# ? Mar 24, 2023 02:50

cinci zoo sniper: Mar 15, 2013

Discendo Vox posted:

Can you provide an example? Because it's normal journalistic practice to include the background/context for new readers past the lede material.

i know it is, theirs was just on the extreme long end of anything ive ever read, and not labelled clearly as background context. i stopped reading them for that reason, as at times it fet like lazy word count padding, with the real article body shorter than the padding

# ? Mar 24, 2023 03:15

Mellow Seas: Oct 9, 2012; Can't post for 10 years!

More like the New York Wastin' My Times amirite

# ? Mar 24, 2023 05:02

Discendo Vox: Mar 21, 2013; We don't need to have that dialogue because it's obvious, trivial, and has already been had a thousand times.

cinci zoo sniper posted:

i know it is, theirs was just on the extreme long end of anything ive ever read, and not labelled clearly as background context. i stopped reading them for that reason, as at times it fet like lazy word count padding, with the real article body shorter than the padding

The new material in articles is often just a paragraph, and it's part of the overall contextual material to avoid failing to provide a clear picture to the reader, so I don't think it's ever labeled as separate. Again, that's normal journalism, practice for like a century.

Discendo Vox fucked around with this message at 06:32 on Mar 24, 2023

# ? Mar 24, 2023 06:27

cinci zoo sniper: Mar 15, 2013

Discendo Vox posted:

The new material in articles is often just a paragraph, and it's part of the overall contextual material to avoid failing to provide a clear picture to the reader, so I don't think it's ever labeled as separate. Again, that's normal journalism, practice for like a century.

I'm a Journalism Enthusiast reading quite a bit of it over the years, if less so focusing on American press specifically, and NYT's practice of copy pasting meaningless volumes of text was incredibly obvious, jarring, and unusual to my experience. Consider this random longer article from Latvian news � scroll to the bottom of it:

quote:

KONTEKSTS:

Pēc "Apvienotā saraksta" iebildēm valdo�ā koalīcija vienojusies no nākamā gada bud�eta likumprojektu paketes izņemt Izglītības likuma grozījumus, kas būtu kalpoju�i kā sākums skolu tīkla reformai. "Apvienotā saraksta" pārstāvju ieskatā �ie grozījumi būtu apdraudēju�i mazo lauku skolu turpmāku pastāvē�anu.

Sarunas par skolu slēg�anu ierasti bijis ļoti sensitīvs temats. To nevēlas pieļaut vietējie iedzīvotāji, un līdz �im pret to parasti ļoti strikti iebilda arī vietējās pa�valdības.

Taču, �ķiet, pēc novadu reformas spēles noteikumi ir mainīju�ies � pa�as vietvaras rūpīgi sāku�as plānot, kā skolu tīklu optimizēt. Eksperti to vērtē kā loģisku lietu kārtību, jo novadu jaunās aprises prasa pārplānot arī skolu tīklu.

This is the normal that I'm used to. NYT's practice, comparatively, could very well be a dark pattern specifically designed to maximize the time I spend on the website, as far as I am concerned. Like, I cannot overstate this - my news budget is hundreds of dollars per year, and subscribing to NYT was an absolutely crazy change for me in this regard, and had me unsubbing promptly thereafter.

# ? Mar 24, 2023 11:05

Discendo Vox: Mar 21, 2013; We don't need to have that dialogue because it's obvious, trivial, and has already been had a thousand times.

I...think you're comparing apples to oranges here; coverage of small new ledes in an ongoing event with a lot of prior context (like a war) is going to look very different from reporting out a new, relatively localized planning announcement. And it's not particularly common to have a literal "context" section in an article (in part because it means people don't read it). I can think of plenty of SEO-screwed lovely journalism examples, but they don't tend to look like what you describe. Could you give an example of the "copy pasting meaningless volumes of text" at the NYT? On a quick review of their material in different formats, I'm having trouble locating articles that match that description.

Discendo Vox fucked around with this message at 12:18 on Mar 24, 2023

# ? Mar 24, 2023 12:12

cinci zoo sniper: Mar 15, 2013

Discendo Vox posted:

I...think you're comparing apples to oranges here; coverage of small new ledes in an ongoing event with a lot of prior context (like a war) is going to look very different from reporting out a new, relatively localized planning announcement. And it's not particularly common to have a literal "context" section in an article (in part because it means people don't read it). I can think of plenty of SEO-screwed lovely journalism examples, but they don't tend to look like what you describe. Could you give an example of the "copy pasting meaningless volumes of text" at the NYT? On a quick review of their material in different formats, I'm having trouble locating articles that match that description.

I don't think I'm coming through clearly enough. I mainly consume non-American journalism (read: at most 2 publications at any given time), and in my media diet context as a rule is concise and non-redundant, and very often explicitly signposted. The crucial difference here is that the journalism I read normally doesn't assume that you need a detailed account of literally everything that has happened for the context. Only some basic information that clearly identifies which conversation the piece belongs to so that you can research it further if you'd like. NYT's reporting on the war, for contrast, had these enormous wall of text woven into every article, regurgitating basic information in every single article every day. Trying to read a few of those per day with the goal of identifying new information would just drive me insane.

The example I referenced, for your context, no pun intended, is not some new procedural minutiae, but a yet another piece in a long-running debate about reforming the public school network, that's been going on for a year. If you insist though that the war is a special topic, here's the copy-paste context used for war articles:

quote:

KONTEKSTS:

2022. gada 24. februārī Krievijas prezidents Vladimirs Putins pavēlēja sākt Krievijas karaspēka iebrukumu Ukrainā. Putins apgalvoja, ka NATO gatavojas izmantot Ukrainu kā placdarmu agresijai pret Krieviju, lai gan �iem apgalvojumiem nebija nekādu pierādījumu.

Ukraina uzskata, ka Putina patiesais mērķis ir iznīcināt Ukrainas valstiskumu un pakļaut �o teritoriju Maskavas kontrolei.

Krievija ar raķetēm regulāri ap�auda Ukrainas pilsētas, lai atstātu ukraiņus bez elektrības un siltuma. Eiropas Parlaments ir pasludinājis Krieviju par terorismu atbalsto�u valsti.

2023. gada 17. martā Starptautiskā Krimināltiesa izdeva orderi Putina arestam, apsūdzot viņu kara noziegumos saistībā ar ukraiņu bērnu deportāciju.

My translation:

quote:

CONTEXT:

On the 24th of February 2022, Vladimir Putin, the President of Russia, did order to launch the incursion of the Russian army into Ukraine. Putin claimed that NATO is preparing to use Ukraine as a bridgehead for aggression against Russia, despite these arguments lacking any evidence.

Ukraine considers destroying Ukrainian statehood, and making the territory subordinate to Moscow's control, to be Putin's true goal.

Russia regularly subjects Ukrainian cities to rocket bombardment, to deprive Ukrainians of electricity and heating. The European Parliament has declared Russia a state sponsor of terrorism.

On the 17th of March 2023, the International Criminal Court did issue an arrest warrant for Putin, accusing him of war crimes in relation to the deportation of Ukrainian children.

# ? Mar 24, 2023 12:54

cinci zoo sniper: Mar 15, 2013

or, more fittingly for the chat thread, my argument is that nyt writes articles as if for idiots. regardless of that being the norm for american press, i ain't got time for it

# ? Mar 24, 2023 12:55

Mellow Seas: Oct 9, 2012; Can't post for 10 years!

Idiots deserve to be informed too.

# ? Mar 24, 2023 13:19

Discendo Vox: Mar 21, 2013; We don't need to have that dialogue because it's obvious, trivial, and has already been had a thousand times.

Okay, so can you give me an example of these articles for idiots, because it now just sounds like you're objecting to providing larger amounts of contextual information.

# ? Mar 24, 2023 13:42

Turgid Flagella: Mar 18, 2023

Discendo Vox posted:

Okay, so can you give me an example of these articles for idiots, because it now just sounds like you're objecting to providing larger amounts of contextual information.

Nice vape cloud bro

# ? Mar 24, 2023 14:02

Paracaidas: Sep 24, 2016; Consistently Tedious!

Discendo Vox posted:

I...think you're comparing apples to oranges here;

At a glance they've eased up their mandate since acquiring The Athletic but one of the early hallmarks of the Times ownership regime was initially much the same as czs describes on the ~~mother~~GrayLadyship

For many breakingish news stories, it was a simple few sentences related to the headline, and then many copy/pasted paragraphs of subheadered "background" that'd be the same under all related headlines and in no way reference or be shaped by the lone graf atop the article.

To paraphrase around an example because their search function is garbage so I can't grab the actuals from a year back:

Panthers trade up for Number 1 pick, here's what you need to know posted:

The Carolina Panthers have traded away their number 9 pick, 2 second rounders, a future first rounder, and wide receiver DJ Moore for the first pick of the 2023 NFL Draft

Dry, but fine. Over time, the relevant beat writers may add single graf "what this means for" the teams and players involved, if they don't do full stories (50/50 those stories are added as links/further reading the original). What would then follow would be the moron's guide to the NFL draft... which is a choice for a strictly paywalled website whose entire brand is elevated commentary, insights, and reporting on sports

95% of the body for every NFL draft 'headline' posted:

What is the NFL Draft?
An annual event where college players are selected by Pro Teams [+2 fluff sentences]

Can my team pay more money to get the better players?
No, because the draft order is based on the team record from the prior season [+4 fluff sentences]

Can my team draft high school players?

Do players have any say where they go?

How long is the draft?

And so on. It'd be common to see the idiot's guide 5 or 6 times a week depending on how often more relevant news bubbled up to deserve a few lines of its own

# ? Mar 24, 2023 16:48

Adbot: ADBOT LOVES YOU

# ? May 21, 2024 14:46

cinci zoo sniper: Mar 15, 2013

Discendo Vox posted:

Okay, so can you give me an example of these articles for idiots, because it now just sounds like you're objecting to providing larger amounts of contextual information.

i am objecting to opaquely mixing short articles with much longer amounts of trivial contextual information. i don't need to be reminded that kyiv is in ukraine and that it is threatened by russia, in half a dozen paragraphs or more. if you read one article every 3 months it's fine, if you read 3-5 articles per day on the subject from the journal for a month straight it's straight up maddening to me. this is a very nyt-specific problem too, since i don't have complaints like that with, e.g., axios or the intercept

# ? Mar 24, 2023 17:01

The Something Awful Forums > Discussion > Debate & Discussion > d&d chat: void scream

«‹›602 »