I'd like to point out that I chose to share this article from AI because it was free of paywalls and ads, objective in its approach, and because AI-generated content is considered public domain.
I'd like to point out that I chose to share this article from AI because it was free of paywalls and ads, objective in its approach, and because AI-generated content is considered public domain.
I think it is still a mistake to post this, because it is wrong in subtle ways that you can't discover without reading all the sources, which defeats the point of reading the AI summary in the...
Exemplary
I think it is still a mistake to post this, because it is wrong in subtle ways that you can't discover without reading all the sources, which defeats the point of reading the AI summary in the first place.
For example, look at the "Organizational Culture Overhaul" section. This section is in the present tense, communicating the idea that Samsung is currently undergoing this overhaul, and heavily implying that it's at least partly in response to their failure to capitalize on the AI boom. It cites 5 sources, but all the content really comes from this one, an article from 2021: https://www.koreatimes.co.kr/www/tech/2024/09/129_318975.html. I think it may have been fooled by the URL, which says 2024, but it seems that the site just re-uploaded it for some reason. The original is https://www.koreatimes.co.kr/www/tech/2021/11/129_318975.html. Perplexity says:
The company is moving away from its traditional hierarchical system towards a more agile and innovation-centered approach. Key changes include:
Simplifying job rankings from seven to four career levels
But that's a bad interpretation of the source article. It actually says:
In Korea, not only Samsung but also many other companies had maintained a seniority and rank-based hierarchy, but they are focusing on establishing a horizontal organizational culture that evaluates individual employees' abilities and rewards them accordingly. Samsung fine-tuned its HR system once before in 2016, simplifying its job ranking from seven notches to four notches ― career level 1 through career level 4.
So they are augmenting the hierarchical system, not necessarily moving away from it, and the job ranking change happened in 2016. Perplexity's present-tense framing of that point is wrong.
Perplexity also often copy-pastes a fragment of a sentence from a source article, but cuts out useful information that a competent human author would have included. For example: "with SK Hynix already supplying HBM3E 8-layer products to Nvidia and planning to mass-produce HBM3E 12-layer products." The source says "SK Hynix is already supplying HBM3E 8-layer products to Nvidia and plans to mass-produce HBM3E 12-layer products in the third quarter for shipment in the fourth quarter." So when Perplexity says they are "planning to mass-produce", that actually was supposed to happen in the third quarter, so it should have already happened and they should be shipping by now. A correct summary would include the dates and/or rephrase the sentence to not say "planning".
One more example of this - Perplexity says: "This advancement has narrowed the operating profit gap between SK Hynix and Samsung to a mere 5.5 billion won in the first half of 2024." Source says: "The operating profit gap between Samsung Electronics and SK Hynix has narrowed to a mere 5.5 billion won (approximately $4.0 million) for the first half of the year." Why cut out the currency conversion? That's useful.
Two more and then I'm done.
Perplexity: "However, Samsung still faces challenges in obtaining approvals for its latest advanced HBM chips".
Source: "The first challenge for Samsung is to get its HBM chips approved by Nvidia"
That is not the same usage of "challenge", Perplexity's usage implies they are already trying and it isn't going well. The source just says it's a difficult thing they'll need to do.
Perplexity: "The intensifying competition has led both companies to aggressively pursue talent acquisition and technological advancements in the HBM sector."
Source: "Samsung must now review its organizational culture and processes, Jun said — echoing previous comments about the need for fundamental change at one of Korea’s oldest companies. It’s begun laying off workers in Southeast Asia, Australia and New Zealand as part of a plan to reduce global headcount by thousands of jobs, Bloomberg News reported last week."
Not sure where it got the "talent acquisition" idea, it's not present in either of the cited sources and one of them directly contradicts it.
Maybe these points all seem minor, but if you are consistently reading AI news this stuff will really add up, and you will be misinformed. Also, all the quality news sources that Perplexity is relying on will either go out of business or start blocking Perplexity, so the AI summaries will gradually get worse.
Is there is a Creative Commons/Public Domain or allowed to republish article reporting into this? I am asking because I don't want to support the broken copyright system, by publishing a...
Is there is a Creative Commons/Public Domain or allowed to republish article reporting into this?
I am asking because I don't want to support the broken copyright system, by publishing a copyrighted articles (with very few exceptions).
I'm not a huge fan of the copyright system either, nor of paywalls, but good quality journalism matters, and that costs money. So I would strongly recommend considering just posting a paywalled...
I'm not a huge fan of the copyright system either, nor of paywalls, but good quality journalism matters, and that costs money. So I would strongly recommend considering just posting a paywalled article if you can't find a similarly high quality paywall free alternative. I typically post paywall free archive.is mirrors to all paywalled articles posted here anyways. Or you can link to a mirror in the comments yourself if you're so inclined, but it's not required or expected.
Wait what? No it is not? Or rather, the debate is still out (although the cat pretty much out of the bag) what the situation is given training material and all that.
and because AI-generated content is considered public domain.
Wait what? No it is not? Or rather, the debate is still out (although the cat pretty much out of the bag) what the situation is given training material and all that.
@Advanced might be thinking of the US Copyright Office's recent-ish interpretation that: Only works created by a human being are protected by copyright. Works created "without human intervention",...
@Advanced might be thinking of the US Copyright Office's recent-ish interpretation that:
Only works created by a human being are protected by copyright.
Works created "without human intervention", such as by artificial intelligence systems or animals, don't have any such protection.
The Copyright Office's page on AI has a bunch of reports and studies that are interesting reads on the subject:
Could be. In this case, we are talking about summarizing content of other news sources. Although citing articles of other news outlets also is something that is generally accepted, so for all...
Could be. In this case, we are talking about summarizing content of other news sources. Although citing articles of other news outlets also is something that is generally accepted, so for all intents and purposes I guess this could be considered public domain.
I still feel like it is an overly simplistic view of the interpretation. While technically, the generated content might be “public domain” if someone would create “The New York Times Rewritten by AI”. And offered all articles from the New York Times after being rewritten by an AI the content might be public domain, but they likely would also be shut down fairly quickly.
I guess what I am getting at is that using “it is AI generated, therefore public domain” might be technically correct. However, that doesn't make it a correct justification for using that content. So it struck me as a very odd justification tacked on to the other ones.
At least in the US (head's up, direct PDF link), no AI-generated work can be copyrighted. While any creative work automatically has copyright granted to the author, AI-authored work has no...
At least in the US (head's up, direct PDF link), no AI-generated work can be copyrighted. While any creative work automatically has copyright granted to the author, AI-authored work has no legal/human author, and so has no copyright. And if it has no copyright, well, that's what the public domain is.
Whether or not AIs themselves are violating copyright with their training data is another question. But on the "is AI-generated content public domain" front, while things may change, they are currently clear.
What you’re saying just isn’t accurate. Im sorry, I don’t mean to single out your post, but I’m getting really frustrated seeing this repeated as fact every few days here on Tildes. From the...
What you’re saying just isn’t accurate. Im sorry, I don’t mean to single out your post, but I’m getting really frustrated seeing this repeated as fact every few days here on Tildes.
From the document you linked:
This policy does not mean that technological tools cannot be part of the creative process. Authors have long used such tools to create their works or to recast, transform, or adapt their expressive authorship. For example, a visual artist who uses Adobe Photoshop to edit an image remains the author of the modified image, and a musical artist may use effects such as guitar pedals when creating a sound recording. In each case, what matters is the extent to which the human had creative control over the work's expression and “actually formed” the traditional elements of authorship.
How much creative control needs to be exercised by the human? According to precedent, hitting the button on a camera is sufficient. The camera isn’t the author, the person who pressed the button is.
Logically, it follows that the AI isn’t the author, the person sitting at the computer is, and the guidance explicitly allows for that. The field’s still new and the details are still being litigated - the precedent for exactly where the boundaries are doesn’t exist yet.
At most, you can say that the (non-binding) guidance suggests they’re being a bit harsher on the amount of human input needed when using a neural net compared to the amount needed when taking a photo. That’s a far, far cry from the computer being the author and all AI work being public domain.
And for what it’s worth, I actually want to see the copyright office being harsher! Copyright is wildly out of control and unfit for purpose in a way that hurts the public and the majority of creators in order to protect the profits of publishers. I just think accuracy is important here.
I don't disagree at all that it logically follows from past rules, but that's not where things are actually at, now, in practice. Without copyright, all* AI work (maybe it'd be more accurate to...
I don't disagree at all that it logically follows from past rules, but that's not where things are actually at, now, in practice.
Without copyright, all* AI work (maybe it'd be more accurate to say "output"?) is public domain, that's some distance behind the front lines of the litigation currently. But you are right that I did neglect to mention that sufficient human input can pull it back out of the public domain. I still feel justified in my earlier reply though. To say flatly that AI generated work is not in the public domain like GP did is substantially more inaccurate than saying that AI output is not copyrightable.
* Since we're talking about accuracy now, it's probably also worth mentioning that there's always exceptions. For example, the output of an AI model can be copyrighted if it's ingested or has otherwise been fed copyrighted creative works.
If you read the full document, even that rejection makes it very clear that they’re considering the balance of human input - they specifically talk about it being an issue when the machine...
If you read the full document, even that rejection makes it very clear that they’re considering the balance of human input - they specifically talk about it being an issue when the machine receives “solely a prompt”. They actually explicitly say they’re only considering the Midjourney part of the toolchain, not the Gigapixel AI part, when coming to their decision - they made a call on Midjourney, not on AI as a whole.
That’s what’s getting to me here, they said that this Midjourney output isn’t acceptable because a prompt isn’t enough to confer authorship. They made comments about the balance of human input, and released guidance saying that some level of technical tool use is acceptable. They even explicitly said that they aren’t ruling either way on the other AI tool used in the work they rejected. But everyone’s reading that as a blanket ruling saying “AI output can’t be copyrighted”.
In the general case, they’ve made one or two narrowly defined decisions that outright say they’re based on the level of input used for a particular work and tool, not on the field as a whole. Those decisions are also inconsistent with other formats and haven’t yet been challenged in court, but that’s almost a side note in my mind - the constraints in their own description of what they’ve ruled on is much more relevant.
In this specific case, we’re talking about text output that actively acknowledges its inputs as (copyrighted) articles from Bloomberg and The Register - one hell of a lot more than the “solely a prompt” that rejection is based on.
It sounds like we’re on the same page about that last bit, but I think it’s so much more than a footnoted exception that it’s genuinely misleading to make blanket statements about AI output as a whole based on a ruling about Midjourney’s particular choice of text to image UX.
Please don't submit something like this again. AI-generated articles are low quality and often contain significant errors and inaccuracies (as @burkaman demonstrated here). They aren't an...
Please don't submit something like this again. AI-generated articles are low quality and often contain significant errors and inaccuracies (as @burkaman demonstrated here). They aren't an appropriate source for news.
I hadn’t really thought about how that works in Korea, but I guess I’m not surprised that their hierarchy is/was encoded into their IT systems. I wonder what kind of real world effects that has on...
Mandating the use of honorific language regardless of position or seniority
Removing rank and seniority information from internal systems
I hadn’t really thought about how that works in Korea, but I guess I’m not surprised that their hierarchy is/was encoded into their IT systems. I wonder what kind of real world effects that has on collaboration.
I know nothing of South Korean culture so I could be way off base, but my experience working in tech in the US is that stuffy formality like this stifles innovation by imposing unnecessary...
Mandating the use of honorific language regardless of position or seniority
I know nothing of South Korean culture so I could be way off base, but my experience working in tech in the US is that stuffy formality like this stifles innovation by imposing unnecessary personal barriers and making work feel less "fun" in general.
From one of the linked sources: I'm not Korean, but my interpretation from just reading this article is that the changes are intended to put people on more equal footing. When you address another...
From one of the linked sources:
In 2017, we reorganized the hierarchy of seven ranks into a more horizontal structure of four Career Levels (CLs) and introduced equal titles for addressing employees of all levels including “nim” (a genderagnostic title of respect), “pro” (short form of “professional”), and English names, in pursuit of a more flexible and inclusive organizational culture. In 2022, we made it mandatory to use the honorific form of language at work at all times regardless of position, title, seniority, or age, and removed information indicating the ranks and seniority of employees from our internal system.
I'm not Korean, but my interpretation from just reading this article is that the changes are intended to put people on more equal footing. When you address another person in Korean, there are several levels of formality. You would generally be expected to use a more formal levels of address for someone who is above you in the company (even outside your reporting tree) and be allowed to use a less formal level when speaking to someone below you. It sounds like they're mandating that people use the same level of formality when addressing others in an attempt to make the culture feel less hierarchical.
Oh joy, the massive fortune 500 company only made 9 trillion dollars in its return instead of 11 trillion. I'm sure glad we live in a society that incentivizes constant shareholder growth or else...
Oh joy, the massive fortune 500 company only made 9 trillion dollars in its return instead of 11 trillion. I'm sure glad we live in a society that incentivizes constant shareholder growth or else you face dissolution. Capitalism really does encourage freedom through the free market and everyone is better for it!
I don't understand this response. For one, the apology is part of a different cultural background; US companies rarely issue "apologies". Secondly, missing net profit expectations by nearly half...
I don't understand this response. For one, the apology is part of a different cultural background; US companies rarely issue "apologies". Secondly, missing net profit expectations by nearly half is pretty bad. If you expected to make $110k this year, and made $68k, even if you're not being imminently foreclosed on, that's a pretty big bummer.
Moreover, it's about trends. Samsung Electronics has been falling behind competitors like TSMC in silicon fabrication, as well as domestic rivals like SK Hynix for memory production. That's bad for the company.
I love it! More money! More expansion! More profits! If not, get outta the economy! I am not satisfyingly contributing to the conversation in here at all, zero expertise, I would just like to say...
I love it! More money! More expansion! More profits! If not, get outta the economy!
I am not satisfyingly contributing to the conversation in here at all, zero expertise, I would just like to say I am struggling to make rent and am that I really hate the landscape of techno-futurism—yes, because I don't understand it, but also because I shouldn't have to understand it to know that its been built on tiny aching fingers. I'm just disassociated and broke and I don't want to have to start a business to define my worth and live in dignity
I'd like to point out that I chose to share this article from AI because it was free of paywalls and ads, objective in its approach, and because AI-generated content is considered public domain.
I think it is still a mistake to post this, because it is wrong in subtle ways that you can't discover without reading all the sources, which defeats the point of reading the AI summary in the first place.
For example, look at the "Organizational Culture Overhaul" section. This section is in the present tense, communicating the idea that Samsung is currently undergoing this overhaul, and heavily implying that it's at least partly in response to their failure to capitalize on the AI boom. It cites 5 sources, but all the content really comes from this one, an article from 2021: https://www.koreatimes.co.kr/www/tech/2024/09/129_318975.html. I think it may have been fooled by the URL, which says 2024, but it seems that the site just re-uploaded it for some reason. The original is https://www.koreatimes.co.kr/www/tech/2021/11/129_318975.html. Perplexity says:
But that's a bad interpretation of the source article. It actually says:
So they are augmenting the hierarchical system, not necessarily moving away from it, and the job ranking change happened in 2016. Perplexity's present-tense framing of that point is wrong.
Perplexity also often copy-pastes a fragment of a sentence from a source article, but cuts out useful information that a competent human author would have included. For example: "with SK Hynix already supplying HBM3E 8-layer products to Nvidia and planning to mass-produce HBM3E 12-layer products." The source says "SK Hynix is already supplying HBM3E 8-layer products to Nvidia and plans to mass-produce HBM3E 12-layer products in the third quarter for shipment in the fourth quarter." So when Perplexity says they are "planning to mass-produce", that actually was supposed to happen in the third quarter, so it should have already happened and they should be shipping by now. A correct summary would include the dates and/or rephrase the sentence to not say "planning".
One more example of this - Perplexity says: "This advancement has narrowed the operating profit gap between SK Hynix and Samsung to a mere 5.5 billion won in the first half of 2024." Source says: "The operating profit gap between Samsung Electronics and SK Hynix has narrowed to a mere 5.5 billion won (approximately $4.0 million) for the first half of the year." Why cut out the currency conversion? That's useful.
Two more and then I'm done.
That is not the same usage of "challenge", Perplexity's usage implies they are already trying and it isn't going well. The source just says it's a difficult thing they'll need to do.
Not sure where it got the "talent acquisition" idea, it's not present in either of the cited sources and one of them directly contradicts it.
Maybe these points all seem minor, but if you are consistently reading AI news this stuff will really add up, and you will be misinformed. Also, all the quality news sources that Perplexity is relying on will either go out of business or start blocking Perplexity, so the AI summaries will gradually get worse.
Is there is a Creative Commons/Public Domain or allowed to republish article reporting into this?
I am asking because I don't want to support the broken copyright system, by publishing a copyrighted articles (with very few exceptions).
I'm not a huge fan of the copyright system either, nor of paywalls, but good quality journalism matters, and that costs money. So I would strongly recommend considering just posting a paywalled article if you can't find a similarly high quality paywall free alternative. I typically post paywall free archive.is mirrors to all paywalled articles posted here anyways. Or you can link to a mirror in the comments yourself if you're so inclined, but it's not required or expected.
Wait what? No it is not? Or rather, the debate is still out (although the cat pretty much out of the bag) what the situation is given training material and all that.
@Advanced might be thinking of the US Copyright Office's recent-ish interpretation that:
The Copyright Office's page on AI has a bunch of reports and studies that are interesting reads on the subject:
https://www.copyright.gov/ai/
Could be. In this case, we are talking about summarizing content of other news sources. Although citing articles of other news outlets also is something that is generally accepted, so for all intents and purposes I guess this could be considered public domain.
I still feel like it is an overly simplistic view of the interpretation. While technically, the generated content might be “public domain” if someone would create “The New York Times Rewritten by AI”. And offered all articles from the New York Times after being rewritten by an AI the content might be public domain, but they likely would also be shut down fairly quickly.
I guess what I am getting at is that using “it is AI generated, therefore public domain” might be technically correct. However, that doesn't make it a correct justification for using that content. So it struck me as a very odd justification tacked on to the other ones.
At least in the US (head's up, direct PDF link), no AI-generated work can be copyrighted. While any creative work automatically has copyright granted to the author, AI-authored work has no legal/human author, and so has no copyright. And if it has no copyright, well, that's what the public domain is.
Whether or not AIs themselves are violating copyright with their training data is another question. But on the "is AI-generated content public domain" front, while things may change, they are currently clear.
What you’re saying just isn’t accurate. Im sorry, I don’t mean to single out your post, but I’m getting really frustrated seeing this repeated as fact every few days here on Tildes.
From the document you linked:
How much creative control needs to be exercised by the human? According to precedent, hitting the button on a camera is sufficient. The camera isn’t the author, the person who pressed the button is.
Logically, it follows that the AI isn’t the author, the person sitting at the computer is, and the guidance explicitly allows for that. The field’s still new and the details are still being litigated - the precedent for exactly where the boundaries are doesn’t exist yet.
At most, you can say that the (non-binding) guidance suggests they’re being a bit harsher on the amount of human input needed when using a neural net compared to the amount needed when taking a photo. That’s a far, far cry from the computer being the author and all AI work being public domain.
And for what it’s worth, I actually want to see the copyright office being harsher! Copyright is wildly out of control and unfit for purpose in a way that hurts the public and the majority of creators in order to protect the profits of publishers. I just think accuracy is important here.
I don't disagree at all that it logically follows from past rules, but that's not where things are actually at, now, in practice.
Without copyright, all* AI work (maybe it'd be more accurate to say "output"?) is public domain, that's some distance behind the front lines of the litigation currently. But you are right that I did neglect to mention that sufficient human input can pull it back out of the public domain. I still feel justified in my earlier reply though. To say flatly that AI generated work is not in the public domain like GP did is substantially more inaccurate than saying that AI output is not copyrightable.
* Since we're talking about accuracy now, it's probably also worth mentioning that there's always exceptions. For example, the output of an AI model can be copyrighted if it's ingested or has otherwise been fed copyrighted creative works.
If you read the full document, even that rejection makes it very clear that they’re considering the balance of human input - they specifically talk about it being an issue when the machine receives “solely a prompt”. They actually explicitly say they’re only considering the Midjourney part of the toolchain, not the Gigapixel AI part, when coming to their decision - they made a call on Midjourney, not on AI as a whole.
That’s what’s getting to me here, they said that this Midjourney output isn’t acceptable because a prompt isn’t enough to confer authorship. They made comments about the balance of human input, and released guidance saying that some level of technical tool use is acceptable. They even explicitly said that they aren’t ruling either way on the other AI tool used in the work they rejected. But everyone’s reading that as a blanket ruling saying “AI output can’t be copyrighted”.
In the general case, they’ve made one or two narrowly defined decisions that outright say they’re based on the level of input used for a particular work and tool, not on the field as a whole. Those decisions are also inconsistent with other formats and haven’t yet been challenged in court, but that’s almost a side note in my mind - the constraints in their own description of what they’ve ruled on is much more relevant.
In this specific case, we’re talking about text output that actively acknowledges its inputs as (copyrighted) articles from Bloomberg and The Register - one hell of a lot more than the “solely a prompt” that rejection is based on.
It sounds like we’re on the same page about that last bit, but I think it’s so much more than a footnoted exception that it’s genuinely misleading to make blanket statements about AI output as a whole based on a ruling about Midjourney’s particular choice of text to image UX.
I replied here to hungariantoast which also covers what you are saying.
Please don't submit something like this again. AI-generated articles are low quality and often contain significant errors and inaccuracies (as @burkaman demonstrated here). They aren't an appropriate source for news.
Sure, I will delete this post as well to not spread misinformation here.
I hadn’t really thought about how that works in Korea, but I guess I’m not surprised that their hierarchy is/was encoded into their IT systems. I wonder what kind of real world effects that has on collaboration.
I know nothing of South Korean culture so I could be way off base, but my experience working in tech in the US is that stuffy formality like this stifles innovation by imposing unnecessary personal barriers and making work feel less "fun" in general.
From one of the linked sources:
I'm not Korean, but my interpretation from just reading this article is that the changes are intended to put people on more equal footing. When you address another person in Korean, there are several levels of formality. You would generally be expected to use a more formal levels of address for someone who is above you in the company (even outside your reporting tree) and be allowed to use a less formal level when speaking to someone below you. It sounds like they're mandating that people use the same level of formality when addressing others in an attempt to make the culture feel less hierarchical.
Oh joy, the massive fortune 500 company only made 9 trillion dollars in its return instead of 11 trillion. I'm sure glad we live in a society that incentivizes constant shareholder growth or else you face dissolution. Capitalism really does encourage freedom through the free market and everyone is better for it!
I don't understand this response. For one, the apology is part of a different cultural background; US companies rarely issue "apologies". Secondly, missing net profit expectations by nearly half is pretty bad. If you expected to make $110k this year, and made $68k, even if you're not being imminently foreclosed on, that's a pretty big bummer.
Moreover, it's about trends. Samsung Electronics has been falling behind competitors like TSMC in silicon fabrication, as well as domestic rivals like SK Hynix for memory production. That's bad for the company.
I love it! More money! More expansion! More profits! If not, get outta the economy!
I am not satisfyingly contributing to the conversation in here at all, zero expertise, I would just like to say I am struggling to make rent and am that I really hate the landscape of techno-futurism—yes, because I don't understand it, but also because I shouldn't have to understand it to know that its been built on tiny aching fingers. I'm just disassociated and broke and I don't want to have to start a business to define my worth and live in dignity