|
BlueBlazer posted:Algorithm = Editorialization = Liable for content recommended I agree, at least on a fundamental level. Though I'm not sure a straight "sort content by new" function or a "new" section that strictly just shows all uploads in chronological order would count. But once you attribute something like "here's the best 10 videos" to content sorting it's an objective call even if you just run it by collecting which videos have the most stars/upvotes/thumbsup/etc. Fart Amplifier posted:Google/Youtube/FB/Amazon need it because everyone expects these services to be free and this is how free services are paid for. I think that its the easiest way for they to generate revenue, along with selling user data which is the real product, but I don't think it's the only means.
|
#
?
Oct 10, 2022 19:00
|
|
|
|
| # ? May 24, 2024 20:07 |
|
|
Fart Amplifier posted:Google/Youtube/FB/Amazon need it because everyone expects these services to be free and this is how free services are paid for. Since when is Amazon a free website? It's a store! The whole idea is that people spend money there.
|
|
#
?
Oct 10, 2022 19:04
|
|
|
Crain posted:I agree, at least on a fundamental level. Though I'm not sure a straight "sort content by new" function or a "new" section that strictly just shows all uploads in chronological order would count. But once you attribute something like "here's the best 10 videos" to content sorting it's an objective call even if you just run it by collecting which videos have the most stars/upvotes/thumbsup/etc. The suggestion I like is that the algorithm has to run in response to an explicit user request to qualify for protection. This would preserve basic listing and sorting (the user did click "show me videos", after all) while removing protections from Youtube's serial autoplay and recommendations section, and it could probably be crafted to kill the algorithmic timeline on other sites
|
|
#
?
Oct 10, 2022 19:05
|
|
|
Regardless of how the midterm elections go, it looks like we'll continue to live in interesting times. https://twitter.com/JonahFurman/status/1579513805834715136
|
|
#
?
Oct 10, 2022 19:12
|
|
|
the idea would be that it is not that the youtube algorithm showed you an extremist video that would potentially make it liable. it would be the nature of the algorithm selecting and promoting extremist content, while not warning users of that behavior. like it is fine to sell a thing that explodes as, well, a thing that is intended to and does explode according to its specifications. it is not ok to sell an exploding child's talking stuffed animal. even if the reason they explode is the same! so "most recently uploaded" and "top ten viewed" would not be problematic because, well, they do what they're advertised to do. they're not personalized recommendations from youtube trying to find the extremist content most likely to hook you. this is something that ordinary liability law should be able to handle rather than a 230-like safe harbor.
|
|
#
?
Oct 10, 2022 19:14
|
|
|
BlueBlazer posted:Algorithm = Editorialization = Liable for content recommended A top 10 list that just objectively lists the top 10 most sold items/clicked articles wouldn't be considered editorialization under the Communications Decency Act. Section 230 is a legal structure that only exists in the U.S. The internet still manages to exist outside of the U.S. and section 230 doesn't completely shield a publisher from liability at all. It acts as sort of an anti-SLAPP law for the internet - by allowing quick dismissals of lawsuits that fall under a broad category of protected actions. Providers still get sued all the time and even lose quite a few of those lawsuits - companies have lost suits for allowing other people to publish personal information on their sites. Section 230 just allows defendants to quickly dismiss suits if a judge agrees.
|
|
#
?
Oct 10, 2022 19:17
|
|
|
Ben Sasse is off to a strong start at his new job in Florida. He released a statement today saying he will formally resign his Senate seat and have the Governor appoint a replacement "imminently," but doesn't say exactly when. https://twitter.com/divyadivyadivya/status/1579551060859183106
|
|
#
?
Oct 10, 2022 20:58
|
|
|
Eric Cantonese posted:Regardless of how the midterm elections go, it looks like we'll continue to live in interesting times. Welp. Hopefully if a strike happens it will be seen as "greedy excutives" vs. "the agreived workers." =/
|
|
#
?
Oct 10, 2022 20:58
|
|
|
Twincityhacker posted:Welp. Hopefully if a strike happens it will be seen as "greedy excutives" vs. "the agreived workers." =/ Right now is in general a good time to make a strong labor stand but also this is going to get blasted as unions driving up inflation or shortages.
|
|
#
?
Oct 10, 2022 21:25
|
|
|
Clarste posted:Why the gently caress is anyone defending the recommendation algorithm anyway? Because recommendations algorithms are hugely useful? Just about every book I checked out from my local library for about 20 years straight was based on Amazon's recommendation system, right up until the removed the old recommendation interface and replaced it with the current (useless) system. It turns out that if you have a vast data set and you can identify people with similar tastes, you can make really good targeted recommendations.
|
|
#
?
Oct 10, 2022 21:56
|
|
|
Leon Trotsky 2012 posted:Ben Sasse is off to a strong start at his new job in Florida. It's incredibly obvious that he was hired for this job because of his political leanings and nothing else.
|
|
#
?
Oct 10, 2022 22:26
|
|
|
Clarste posted:Why the gently caress is anyone defending the recommendation algorithm anyway? It's probably a large part of the reason why Sanders' campaign in 2016 was so popular and didn't flop like Kucinich's 2008 campaign/etc 
|
|
#
?
Oct 10, 2022 22:28
|
|
|
Kalit posted:It's probably a large part of the reason why Sanders' campaign in 2016 was so popular and didn't flop like Kucinich's 2008 campaign/etc there have been a lot of conspiracy theories about malevolent outside forces being responsible for Bernie Sanders, but blaming Youtube's a new one
|
|
#
?
Oct 10, 2022 22:42
|
|
|
Kalit posted:It's probably a large part of the reason why Sanders' campaign in 2016 was so popular and didn't flop like Kucinich's 2008 campaign/etc You mean Howard Dean. Kucinich flopped because he's not exactly charismatic (especially compared to Obama, Edwards, even Clinton), and because  was already taking the role of Internet Candidate. Dean had that in 2004 but 2004 internet lacked the penetration to carry a campaign. In 08/12 Ron Paul got far more votes than he otherwise would have thanks to social media presence/spam. His campaign failed because he was Ron Paul and not because his message failed to reach people.
|
|
#
?
Oct 10, 2022 22:42
|
|
|
Yeowch!!! My Balls!!! posted:there have been a lot of conspiracy theories about malevolent outside forces being responsible for Bernie Sanders, but blaming Youtube's a new one I'm talking about social media recommendation algorithms in general, not just youtube 
Kalit fucked around with this message at 22:48 on Oct 10, 2022 |
|
#
?
Oct 10, 2022 22:44
|
|
|
Kalit posted:I'm talking about social media recommendation algorithms in general, not just youtube have you ever wondered why so much hetero pornography, over the last several years, has moved into the genre of Technically Not Incest Winky Face? the easy, obvious, and even technically accurate answer is that The Algorithm likes it better than regular type. oh, those pernicious algorithms, if only they could be changed. simple technical solution to a technical problem, right? obviously not, otherwise I wouldn't be bringing this up. turns out there's a guy fairly high up at Pornhub who is into that kind of thing. and so, when designing the pornhub algorithm, he prioritized the kind of poo poo he liked to see, and who is going to tell that guy no. end result, several years down the line, the market has adapted to suit the preferences of the guy in a position to dictate what it looks like. as a professional algorithm-wrangler (thankfully outside the social media space) let me teach you the third of the great secrets of modern IT. the first two, of course, are "there's no such thing as the cloud, it's just someone else's computer," and "there's no such thing as the blockchain, it's just a bunch of other people's computers you don't get access to." there is no such thing as The Algorithm. it's just a guy. and as with the first two, there is an unbelievable amount of cachet and money in telling people who consider themselves very smart indeed that the word is instead a synonym for 'magic.' the question "why does the algorithm boost these things" always has the same answer. "because the companies boosting them want to see more of them." blaming Donald Trump, Bernie Sanders, or whoever the microcelebrity of the day is on algorithms is an intellectually bankrupt attempt to deflect a political problem into the realm of technology. someone wants to see more of these things. they are the actual problem that must be addressed. asking them to use different algorithms is wiring the Titanic PLEASE ROTATE DECK CHAIRS STOP
|
|
#
?
Oct 11, 2022 00:06
|
|
|
Yeowch!!! My Balls!!! posted:there have been a lot of conspiracy theories about malevolent outside forces being responsible for Bernie Sanders, but blaming Youtube's a new one Media stories RE Bernie's success on social media and the attendant algorithmic discussion (with increasingly accurate caveats) date back to, at latest, February 2016. It's a Slate story about Bernie back in early 2016 so for obvious reasons I'm linking solely as evidence that the "Personalized/algorithmic social media helped Sanders breakthrough" narrative is anything but new.
|
|
#
?
Oct 11, 2022 01:45
|
|
|
Yeowch!!! My Balls!!! posted:have you ever wondered why so much hetero pornography, over the last several years, has moved into the genre of Technically Not Incest Winky Face? Fun Fact: The actual answer to your incest porn question is actually the financial crisis in 2009, declining revenues from porn, and the consolidation of most major porn streaming sites and studios under Feras Antoon. New York magazine actually did an interesting story about how the financial crisis caused the rise of incest porn. But, the short version is that Feras Antoon founded Brazzers.com and was the first major site to push "MILF"-centric porn marketing and targeted niche products. During the financial crisis, he bought YouPorn, Pornhub, RedTube, XTube, and a few other streaming sites/studios. These sites were bleeding money, because so much of the content was free and so few people were paying for porn during the financial crisis, that he tried to make the industry more efficient by merging the studios and streaming sites. After that, he was looking for a new marketing niche and found that the taboo of fake incest drove up engagement and purchases from people in that niche and, even more importantly, cost nothing extra to just have an actor say "Step-Mom" versus the costs of other niche categories had in hiring specific models or production costs. Then, everyone else started to copy it. So, basically incest porn differentiated the videos in a market that was saturated and cost nothing extra to produce - because of capitalism and people bein' nasty. The same guy is also basically responsible for OnlyFans. Because part of his cost-cutting plan was to merge studios and distribution networks and then squeeze the actors on residuals and payments. So, major actors looked for distribution methods outside of the studio system. Leon Trotsky 2012 fucked around with this message at 02:04 on Oct 11, 2022 |
|
#
?
Oct 11, 2022 01:55
|
|
|
Yeowch!!! My Balls!!! posted:have you ever wondered why so much hetero pornography, over the last several years, has moved into the genre of Technically Not Incest Winky Face?  I’m honestly confused on what your point is. Let me be explicitly clear about what my point is: Sanders’ 2016 campaign probably got a lot of help because of social media algorithms. And more importantly, his political priorities got majorly boosted during this time to be accepted as mainstream. Hence, why I think social media algorithms was a good thing in this instance.To explain further, it’s no secret that social media boosted him a lot. Before/early on in this campaign, he wasn’t a high profile name. But during this campaign, you can see a lot of new people learning about him via social media. Now, obviously this is not a 1:1 ratio of social media algorithms influencing this. AFAIK, this aspect hasn’t been explicitly studied. If I’m wrong on that, please correct me. But I think it’s likely that the algorithms helped spread the word of Sanders and his ideas, since social media [obviously] had an extremely positive impact on his campaign. The unspoken assumption I’m making is social media is as popular as it is because of these algorithms. I could probably dig out studies if you want to refute this, but I hope you accept that assumption since we all know that companies want to maximize user engagement Kalit fucked around with this message at 01:59 on Oct 11, 2022 |
|
#
?
Oct 11, 2022 01:55
|
|
|
Leon Trotsky 2012 posted:Fun Fact: The actual answer to your incest porn question is actually the financial crisis in 2009, declining revenues from porn, and the consolidation of most major porn streaming sites and studios under Feras Antoon. Yeah, this is the correct story. But it does just change it from the guy who controls the algorithm pushed incest porn because he's horny to it being because he's greedy. Pornhub is advertising for his productions.
|
|
#
?
Oct 11, 2022 02:01
|
|
|
Kalit posted:
and the assumption I am calling out as incorrect is that 'the algorithm' is something meaningfully distinct from 'the desires of the people running the companies in question.'
|
|
#
?
Oct 11, 2022 02:11
|
|
|
Yeowch!!! My Balls!!! posted:and the assumption I am calling out as incorrect is that 'the algorithm' is something meaningfully distinct from 'the desires of the people running the companies in question.' So you're taking the position that algorithmic sorting functions, including ML-based ones, do not produce effects outside those intended by their designers?
|
|
#
?
Oct 11, 2022 02:17
|
|
|
Discendo Vox posted:So you're taking the position that algorithmic sorting functions, including ML-based ones, do not produce effects outside those intended by their designers? No the opposite, their designers are baking in effects they don't even realize or care to control for.
|
|
#
?
Oct 11, 2022 02:23
|
|
|
Discendo Vox posted:So you're taking the position that algorithmic sorting functions, including ML-based ones, do not produce effects outside those intended by their designers? I think they’re saying that the only meaningful effect as far as the owners are concerned is an increase in the rate of profit via increasing the net amount of viewing time, whether that’s because of Elsa taking Spiderman to the doctor or minecraft streams where they talk about the great replacement for six hours.
|
|
#
?
Oct 11, 2022 02:26
|
|
|
Yeowch!!! My Balls!!! posted:and the assumption I am calling out as incorrect is that 'the algorithm' is something meaningfully distinct from 'the desires of the people running the companies in question.' Obviously, human bias is always a consideration. But, like DV brought up, do you think that the outcome of these algorithms are mostly/completely independent of the users’ interests?
|
|
#
?
Oct 11, 2022 02:26
|
|
|
Discendo Vox posted:So you're taking the position that algorithmic sorting functions, including ML-based ones, do not produce effects outside those intended by their designers? if they produce effects contrary to those intended by their designers, they are promptly retooled to no longer produce them. witness the amazing high-tech polling apparatus of 2016, where any algorithm producing the clearly incorrect result was culled from the herd for being defective, or the abortive efforts of Facebook and Twitter to censor neo-nazi hate speech, shut down in the preliminary stages when it was revealed conservative politicians would run afoul of them. a crude form of natural selection functions here. cosmetic mutations- a little incest porn here, a couple Have You Considered Spending More To Make Your Suicide Attempt More Successful kits there- are permitted to survive, and drift either towards fixation or oblivion. outcomes contrary to the designers' desires does not make it to launch. long story short, the people at Facebook saying you 'we didn't know we were building a news network for fascists, it was just the algorithm' while giving the Daily Wire full editorial control of their platform were lying to you. i know, shocker.
|
|
#
?
Oct 11, 2022 02:34
|
|
|
Yeowch!!! My Balls!!! posted:a crude form of natural selection functions here. cosmetic mutations- a little incest porn here, a couple Have You Considered Spending More To Make Your Suicide Attempt More Successful kits there- are permitted to survive, and drift either towards fixation or oblivion. Just to be completely clear, you think selling suicide kits is the desire of the person or people who created the system?
|
|
#
?
Oct 11, 2022 02:39
|
|
|
I've gotten several different interpretations here and not much of a clear statement. Are you saying, for example, the "suicide kit" product recommendations at amazon weren't contrary to the designs of the people at amazon?
|
|
#
?
Oct 11, 2022 02:40
|
|
|
DeadlyMuffin posted:Just to be completely clear, you think selling suicide kits is the desire of the person or people who created the system? I mean a profit maximizing algorithm doesn't care what it is selling or to whom. It's not a state they specifically sought-after but it is absolutely a desired outcome under the parameter space their algorithm is optimizing for. Heck Yes! Loam! fucked around with this message at 02:50 on Oct 11, 2022 |
|
#
?
Oct 11, 2022 02:47
|
|
|
Discendo Vox posted:I've gotten several different interpretations here and not much of a clear statement. Are you saying, for example, the "suicide kit" product recommendations at amazon weren't contrary to the designs of the people at amazon? There’s a category error going on here. When people say that amazon recommending suicide kits is consistent with the designs of amazon, they mean that it makes money for amazon and that the entire design of amazon is to make money. They mean that amazon is entirely indifferent to anything other than making money. That’s different from saying that amazon desires to kill people, though the effect is the same. You could ask the same thing by wondering if Union Carbide intended to kill people.
|
|
#
?
Oct 11, 2022 02:50
|
|
|
Discendo Vox posted:I've gotten several different interpretations here and not much of a clear statement. Are you saying, for example, the "suicide kit" product recommendations at amazon weren't contrary to the designs of the people at amazon? correct. they considered the outcome value-neutral at worst. they were making extra money by selling customers additional products other people wanted to buy-system functioning as designed, and as the gun companies have successfully argued time and time again, it is not their problem what the products get used for. if noone outside the company had made a stink about it, they gladly would have kept doing it until the sun went dark. this was the algorithm functioning as desired. the apathy of its designers to any potential loss of human life was not a bug, but an indication it was functioning as planned. then outside political pressure adjusted the desires of the algorithm's custodians, and miraculously, something that had been functioning perfectly the day before was suddenly decreed to be faulty.
|
|
#
?
Oct 11, 2022 02:52
|
|
|
Yeowch!!! My Balls!!! posted:correct. they considered the outcome value-neutral at worst. they were making extra money by selling customers additional products other people wanted to buy-system functioning as designed, and as the gun companies have successfully argued time and time again, it is not their problem what the products get used for. if noone outside the company had made a stink about it, they gladly would have kept doing it until the sun went dark. Are you saying that the execs at Facebook/Twitter/etc actively wanted Sanders 2016 campaign to be successful? If that's true, holy poo poo, I just became a huge fan of them! (USER WAS PUT ON PROBATION FOR THIS POST)
|
|
#
?
Oct 11, 2022 02:55
|
|
|
Yeowch!!! My Balls!!! posted:correct. they considered the outcome value-neutral at worst. they were making extra money by selling customers additional products other people wanted to buy-system functioning as designed, and as the gun companies have successfully argued time and time again, it is not their problem what the products get used for. if noone outside the company had made a stink about it, they gladly would have kept doing it until the sun went dark. As presented your argument appears completely tautological. It is unclear that there is any set of facts which could disprove it, because it's asserting the system and its designers' intentions are a unity, and any change to the system can only reflect an instant change in those intentions. Discendo Vox fucked around with this message at 02:58 on Oct 11, 2022 |
|
#
?
Oct 11, 2022 02:56
|
|
|
Kalit posted:Are you saying that the execs at Facebook/Twitter/etc actively wanted Sanders 2016 campaign to be successful? If that's true, holy poo poo, I just became a huge fan of them! This doesn't follow at all from what anyone is claiming. And you should not be a fan of huge media company execs just as a rule of thumb.
|
|
#
?
Oct 11, 2022 02:56
|
|
|
Heck Yes! Loam! posted:This doesn't follow at all from what anyone is claiming. ....it was sarcasm. Based on Yeowch!!! My Balls!!! doubling down on their claim of Yeowch!!! My Balls!!! posted:if they produce effects contrary to those intended by their designers, they are promptly retooled to no longer produce them
|
|
#
?
Oct 11, 2022 02:58
|
|
|
Heck Yes! Loam! posted:I mean a profit maximizing algorithm doesn't care what it is selling or to whom. It's not a state they specifically sought-after but it is absolutely a desired outcome under the parameter space their algorithm is optimizing for. Isn't it more likely that this was an unexpected outcome of how the parameter space was defined rather than selling suicide kits is the desire of the person or people who created the system, which is what YMB is implying? Yeowch!!! My Balls!!! posted:outcomes contrary to the designers' desires does not make it to launch. Suicide kits are an outcome post launch, therefore they must not be contrary to the designers' desires.
|
|
#
?
Oct 11, 2022 02:59
|
|
|
Kalit posted:....it was sarcasm. Based on Yeowch!!! My Balls!!! doubling down on their claim of That still doesn't follow. I completely agree with YMBs perspective on this. As someone with experience with machine learning and Al Gore Rhythms, these outcomes like suicide kits and companies advertising pregnancy products before women even know they are pregnant, are completely predictable. Machine learning isn't magic, and even with modern tooling, controls, testing, and data quality, you get some ridiculous things when you apply it to people. This doesn't mean those things are designed on purpose, or that the designers have any control over those outcomes. In fact quite the opposite. Most often the designers are completely blind to the bias and nefarious application of their algorithm. In this case Amazon was just pairing items that are linked together in a graph because of some previous human that bought them together before because it will sell more of both of the items in question when they are linked. Heck Yes! Loam! fucked around with this message at 03:07 on Oct 11, 2022 |
|
#
?
Oct 11, 2022 03:04
|
|
|
This feels like a very D&D conversation where someone said something generally agreeable, that the algorithms that feed us content have a negative effect on us both by design and accident but in a way that's far too verbose so everyone has found a point to quibble. (USER WAS PUT ON PROBATION FOR THIS POST)
|
|
#
?
Oct 11, 2022 03:12
|
|
|
Discendo Vox posted:As presented your argument appears completely tautological. It is unclear that there is any set of facts which could disprove it, because it's asserting the system and its designers' intentions are a unity, and any change to the system can only reflect an instant change in those intentions. not a unity. cosmetic fluctuations, those the organization does not consider contrary to its aims, are not cause for removal. 2016 polls could claim Texas was going blue and be considered credible! it was only if your system predicted Trump might win that it needed correction before release to the public. it is only results contrary to the designers' intent that do not make it live. results the designers do not care about are free to flourish. this results in problems when something abruptly makes the designers care about something they did not care about before. if you can show an organization making a reversal on a core policy when presented by an algorithmic result contrary to its mission, you will have found a counterexample. it is just that you will be looking for a very long time, because the Iron Law of Institutions is a motherfucker.
|
|
#
?
Oct 11, 2022 03:13
|
|
|
|
| # ? May 24, 2024 20:07 |
|
|
DeadlyMuffin posted:Isn't it more likely that this was an unexpected outcome of how the parameter space was defined rather than selling suicide kits is the desire of the person or people who created the system, which is what YMB is implying? unexpected outcomes like republican politicians being caught up in neo-nazi filters are cause for dismissing an algorithm as unviable prior to launch. build your own suicide kits were not. why would this be the case, if not 'the designers cared about one outcome, and not the other?'
|
|
#
?
Oct 11, 2022 03:16
|
|