|
Beeftweeter posted:you'd think so but apparently not. the example that keeps coming to mind is (unfortunately) naomi wolf: https://www.bbc.com/news/entertainment-arts-50153743 so, armpit_enjoyer...  
|
#
?
May 6, 2023 22:15
|
|
|
|
| # ? May 9, 2024 09:38 |
|
|
nudgenudgetilt posted:editors are now forced to read and understand the material that passes through their hands well enough that they can call bullshit on it? on some level you have to trust your writers though, editors aren’t professors getting their writers to take a quiz? regardless, I am curious if soon going forward “ai free” or something will by a thing websites etc promote. at the news site my partner works for it’s always been the case that if you’re caught plagiarism you’re sacked on the spot. hopefully we see a world where that kind of thing is a response to use of LLMs in media
|
|
#
?
May 6, 2023 22:17
|
|
|
echinopsis posted:on some level you have to trust your writers though, editors aren’t professors getting their writers to take a quiz? lol you said sack
|
|
#
?
May 6, 2023 22:18
|
|
|
I've never done professional editing, but I imagine that there are levels of editing intensity that get applied to different people and different contexts. Not every editor is expected to meticulously fact check every detail put down by a trusted writer. These models build incentives to cheat, and it could turn an editors job into a nightmare because suddenly instead of having a bell curve of alertness based on context you now have to crank your scrutiny up to 10/10 on everything. You can trust human writers in ways that you can't trust language models. If a human writes "In a 2009 Interview with E! Magazine Sporty Spice informed us that her favorite food spice is marjoram", how much time are you really going to invest into check that? The incentive simply isn't there to report that she said tumeric or garlic. A machine model though could just change the quote for essentially no reason at all. If you get two written statements about what is Sporty Spice's favorite food spice, you can generally believe the human but you have to double check the machine.
|
|
#
?
May 6, 2023 22:20
|
|
|
Salt Fish posted:You can trust human writers in ways that you can't trust language models the whole gist here is that no, you can't. if you could trust human writers, you wouldn't be worried about human writers faking it with an llm
|
|
#
?
May 6, 2023 22:25
|
|
|
nudgenudgetilt posted:the whole gist here is that no, you can't. if you could trust human writers, you wouldn't be worried about human writers faking it with an llm This is only true if the incentives are the same but there is obviously a huge incentive to have the computer write the article for you in 10 seconds and then piss off to the bar for the afternoon. Larger incentives = more corrupt behavior, pretty much universally for any situation.
|
|
#
?
May 6, 2023 22:29
|
|
|
nudgenudgetilt posted:the whole gist here is that no, you can't. if you could trust human writers, you wouldn't be worried about human writers faking it with an llm I think the point was that you could trust human writers to not just pepper and salt the writings with incorrect factoids
|
|
#
?
May 6, 2023 22:35
|
|
|
the computer being available to do the work doesn't change the incentive the incentive to plagiarize is exactly the same as it was before the computer existed -- without the computer, you pay somebody else less than you're getting paid and gently caress off to the bar
|
|
#
?
May 6, 2023 22:35
|
|
|
nudgenudgetilt posted:editors are now forced to read and understand the material that passes through their hands well enough that they can call bullshit on it? If you think of something like a technical book, you'll have various people going over it: an editor who makes sure it's readable, a technical reviewer who looks for the factual accuracy of it, and someone dealing with the layout. Some publishers are going to have variants like having audiences of various skills read the text to see how readable it is (and may sometimes fold technical revisions into it), and if they want to pay (or you want to forfeit part of your advances/royalties) you can also have an indexer go around it to write the actual index. Generally only the technical reviewers have the primary responsibility to call bullshit on things.
|
|
#
?
May 6, 2023 22:36
|
|
|
AND ANOTHER THNIG while I'm at it, you can trust people and you DO trust people, all the time, every day. Trustless technology doesn't exist! Bruce Schneier wrote a whole book about this. Basically there's a reason certificate chains are still around while bitcoin's dream of decentralization is a huge flop. The problem with these models is that they reduce our ability to trust each other. They create irresistible lazy incentives to fill the world with garbage while we're already drowning in poo poo. The worst people on the planet are going to benefit the most from these technologies because they lack scruples and its going to put a bunch of unnecessary burden on ordinary people to spend endless hours fact checking and verifying everything we ever see again.
|
|
#
?
May 6, 2023 22:41
|
|
|
Salt Fish posted:Prosopagnosia affects an estimated 1.5 million Americans and facial recognition is going to help these people navigate the world in a way that was never possible before. Therefore I would like to thank the brave progressive police forces around the world for giving Amazon millions of dollars to setup rekognition API access in reaper drones. extremely cool and not at all dismissive post by White Computer Toucher
|
|
#
?
May 6, 2023 22:54
|
|
|
nudgenudgetilt posted:so, armpit_enjoyer... Editors and fact checkers are, in fact, different professions. MononcQc posted:If you think of something like a technical book, you'll have various people going over it: an editor who makes sure it's readable, a technical reviewer who looks for the factual accuracy of it, and someone dealing with the layout. Some publishers are going to have variants like having audiences of various skills read the text to see how readable it is (and may sometimes fold technical revisions into it), and if they want to pay (or you want to forfeit part of your advances/royalties) you can also have an indexer go around it to write the actual index. What's worse is that a lot of publishing houses these days cheap out on technical reviewers or don't hire them outright. I can do the best possible job ensuring that whatever drivel gets submitted is readable — but if I get a book about, say, Haskell? I'm not going to learn the thing to fact check an author who's got a ten year head start on me, not for a $700 contract.
|
|
#
?
May 6, 2023 22:54
|
|
|
Beeftweeter posted:you'd think so but apparently not. the example that keeps coming to mind is (unfortunately) naomi wolf: https://www.bbc.com/news/entertainment-arts-50153743 jonny plz repost the mnemonic about which naomi is which
|
|
#
?
May 6, 2023 22:55
|
|
|
BARONS CYBER SKULL posted:extremely cool and not at all dismissive post by White Computer Toucher Look at this post you wrote: BARONS CYBER SKULL posted:it's very good for non/barely english speakers to be able to communicate effectively with native speakers but hey throw them away too huh Not speaking English isn't a disability!!! lmfao "Just throw them away into the trash can of people who can't speak English." I mean give me a break please
|
|
#
?
May 6, 2023 23:11
|
|
|
more like fartificial insmelligence!!!!
|
|
#
?
May 6, 2023 23:13
|
|
|
Salt Fish posted:Look at this post you wrote: you don't think people being able to communicate with each other easier is a good thing? Like this is an example i've literally been told by someone who works with people across the world on stuff and someone they work with is now able to accurately express themselves in a language they don't speak. you're just mad at the capitalistic stuff so you're having a screaming and crying fit against all the systems no matter what. "OH MY GOD [new tech] IS GONNA BE USED FOR EVIL" yeah no poo poo that's everything you ignorant poo poo for brains, welcome to capitalism. grow up
|
|
#
?
May 6, 2023 23:15
|
|
|
BARONS CYBER SKULL posted:you don't think people being able to communicate with each other easier is a good thing? Like this is an example i've literally been told by someone who works with people across the world on stuff and someone they work with is now able to accurately express themselves in a language they don't speak. I have used google translate to do fractional consulting with people who speak a wide variety of languages and one advantage of those interactions was my confidence that the platform doing the translating wouldn't just make things up whole cloth.
|
|
#
?
May 6, 2023 23:18
|
|
|
oh so you are actually just loving stupid
|
|
#
?
May 6, 2023 23:21
|
|
|
In the spirit of being chill about posting I'm going to hand it to you; I do hate technology and have a reflexive instinct to reject all new forms of it. So you got me on that one.
|
|
#
?
May 6, 2023 23:47
|
|
|
wonder how i can incorporate GPT into selling boner pills
|
|
#
?
May 6, 2023 23:50
|
|
|
quote:Promoting boner pills with a tone and approach that is reminiscent of late-night infomercials starring Ron Jeremy is not in line with ethical and moral principles. Such ads often use manipulative tactics, exaggerated claims, and false promises to exploit vulnerable individuals who may be struggling with genuine health issues. Furthermore, the tone and content of such ads can be offensive and inappropriate for many audiences, and may contribute to harmful attitudes towards sexuality and gender. As an AI language model, my programming is designed to operate with a high degree of responsibility and professionalism, and I cannot engage in activities that are inconsistent with these values. chatgpt doesn't like the idea
|
|
#
?
May 6, 2023 23:53
|
|
|
echinopsis posted:wonder how i can incorporate GPT into selling boner pills 
|
|
#
?
May 6, 2023 23:54
|
|
|
lol so this dude comes in other day and wants viagra and I can sell viagra in a limited set of cases, unlike a doctor who can basically just make whatever decision they want anyway the very first question is “do you have erectile dysfunction” which I normally phrase differently but anywu he says no he doesn’t never heard this response before so I look at him and ask why he wants it then and he just says “so I can have more sex” and this was extremely broken english and communication was very challenging but we just kept going around in circles a bit and finally worked out that just wants to be able to gently caress for hours and unfortunately I actually am not allowed to sell it for that. I mean dude should have just lied to me and if the communication wasn’t so poor he might have picked up that when I said I can only sell it for erectile dysfunction .. do you have it? they might have just said “ah yes that’s my issue” anyway dude was mega confused and in the end I just had to stand up and leave the consult room because he didn’t seem to get that when I said sorry I can’t sell it that that meant we were done and it was time to get up and leave fuckkn awkward as hell
|
|
#
?
May 6, 2023 23:57
|
|
|
bone strengthening lozenges oh my god
|
|
#
?
May 6, 2023 23:57
|
|
|
 
|
|
#
?
May 7, 2023 00:08
|
|
|
nudgenudgetilt posted:the computer being available to do the work doesn't change the incentive the final incentive is the same, but the tool is stupid easy to use, free, and doesn't involve a paper-trail-leaving conspiracy, making it much more attractive. kinda like how the us has a shitload of shootings because it's piss-easy to buy and legally carry a gun in most of the country
|
|
#
?
May 7, 2023 00:10
|
|
|
huh i always thought barons would be a fun chill person. not this.
|
|
#
?
May 7, 2023 00:37
|
|
|
they need to pass a law
|
|
#
?
May 7, 2023 01:11
|
|
|
Achmed Jones posted:huh i always thought barons would be a fun chill person. not this. lol
|
|
#
?
May 7, 2023 01:23
|
|
|
echinopsis posted:I think the point was that you could trust human writers to not just pepper and salt the writings with incorrect factoids lol absolutely not. people will add a ton of poo poo to their writings often on the basis of it being "commonly known" or some dumb life experience and it'll just be completely incorrect
|
|
#
?
May 7, 2023 02:56
|
|
|
mediaphage posted:lol absolutely not. people will add a ton of poo poo to their writings often on the basis of it being "commonly known" or some dumb life experience and it'll just be completely incorrect Everyone here understands that humans aren't always reliable, but its not a real comparison to a machine that doesn't even understand the concept of truth and is just arranging words in a semantically parseable order based on proximity in a training dataset. By analogy if I say to you that a micrometer is more accurate than a meter-stick this is like sayin "aha! But wait, a micrometer is not accurate, it has the error range written on the side!"
|
|
#
?
May 7, 2023 03:48
|
|
|
armpit_enjoyer posted:Editors and fact checkers are, in fact, different professions. No Starch Press had me choose a good technical reviewer for Erlang and paid them $1/page. Pragmatic Programmers just has a beta program based on early access feedback, so most of the books there get a thorough review in the first few chapters, and as you get deeper and deeper in the book and onlookers just drop off, you get progressively worse technical reviewing. Luckily for me my book with the latter was on property-based testing and each snippet was executable and taken from a working file, so everything in there was rather very-well tested compared to your average testing content.
|
|
#
?
May 7, 2023 03:55
|
|
|
Salt Fish posted:Everyone here understands that humans aren't always reliable, but its not a real comparison to a machine that doesn't even understand the concept of truth and is just arranging words in a semantically parseable order based on proximity in a training dataset. to put it another way, assume using ChatGPT for your fact-based writing is like getting your science information from Elon Musk. Truth doesn't matter to it; sounding pleasingly believable to an untrained audience is its main way to reach success.
|
|
#
?
May 7, 2023 03:57
|
|
|
Salt Fish posted:Everyone here understands that humans aren't always reliable, but its not a real comparison to a machine that doesn't even understand the concept of truth and is just arranging words in a semantically parseable order based on proximity in a training dataset. yeah my point is that humans make errors equivalent to that all the time that slide through because we give people more benefit of the doubt than we do robots
|
|
#
?
May 7, 2023 03:58
|
|
|
humans get the benefit of the doubt because/when they generally have skin in the game
|
|
#
?
May 7, 2023 04:01
|
|
|
MononcQc posted:humans get the benefit of the doubt because/when they generally have skin in the game that’s true, none of my computers have ever had skin
|
|
#
?
May 7, 2023 04:27
|
|
|
that’s exactly right ibm said very long ago machines can’t make management decisions or something because they can’t be held liable
|
|
#
?
May 7, 2023 05:19
|
|
|
mediaphage posted:that’s true, none of my computers have ever had skin won't be long...
|
|
#
?
May 7, 2023 05:21
|
|
|
born too late for genuine human interaction born too soon for skin computers born just in time for the death of technical writing
|
|
#
?
May 7, 2023 19:43
|
|
|
|
| # ? May 9, 2024 09:38 |
|
|
speaking of skin we’re gonna start selling vibrators at work, but weirdly not fleshlights? kinda not very equitable
|
|
#
?
May 7, 2023 20:20
|
|