Firstly the lawsuit alleges that the AI 'read' an illegal copy on the web, not a legally bought copy from a bookshop. Maybe OpenAI will disprove this by pointing to the kindle receipts or whatever, but somehow I doubt that, it's not the techbro way.
Secondly, you're implying that it's OK to download the work one word/sentence/page at a time, read it, discard it, then do the same for the next one. We have a word for that, it's called streaming.
We know it is legal to stream a movie from netflix/amazon/apple/whoever. But streaming an illegal copy is still illegal, so the underlying source of the material is important.
Thirdly I would like a source for the assertion that AI models use streaming to ingest data. It seems to me that by using streaming would be incredibly inefficient. Every time you tweak the model and want to see how it does, you'd need to scan the whole internet again. If you want to control what goes into the model, you need to be able to review the data. Data storage is cheap, and the only thing you need to store is text, which is highly compressible.
Does Twitter have an undelete post function? I'm wondering why they need to keep deleted posts. Even if you need to keep a record that the post happened, you could save space by deleting the contents.
And all this because Elon got suckered into massively overpaying for the company because he didn’t understand literally anything about social media, and then saddled the company with a huge, unnecessary debt, and his way to deal with that was to put people at risk?
Do you genuinely believe that Elon would act differently if he paid half as much for the platform?
I do not.
How does "Bungie have plenty of options to deal with cheaters and cheat purveyors" when it can't get the defendant to show up in court, nor collect the judgement against them?
I think copyright is the wrong paradigm to look at this case, unless you can prove that they have a copy of your work on their training servers (which I assume they do).
The angle to attack this is licencing. Any art that is explicitly posted under a licence with a Non-commercial clause can't be used by a commercial AI model for training, because the company is using it for a commercial project. The CC BY-NC-SA would be an example of that https://creativecommons.org/licenses/by-nc-sa/4.0
Whether the AI generates an exact copy or similar style is irrelevant because commercial usage is a breach of the licence terms.
I suspect any art posted without a specific licence can probably do the same because logically the creator retained their rights, including whether it could be used for commercial projects.
DeviantArt might have a terms of service argument against this, for art posted on their site, but not art scraped from elsewhere. And OpenAI may have an argument they are non-profit, but wikipedia seems to indicate they aren't just a non-profit anymore.
Yeah, the government under Boris made some clear efforts to influence the actual output of the BBC, and in some cases apparently succeeded.
https://www.theguardian.com/media/2023/mar/15/troubling-messages-showing-no-10-pressure-on-bbc-need-investigating-say-former-staff
While the ruling is interesting, the real test is whether it survives appeal. Courts can and do make mistakes, and both the Court of Appeal and Supreme Court of Kenya could overrule this decision.
Also because it's issued by a specialised court, I suspect that this ruling would have no weight at other courts within Kenya.
The current scandals do NOT show anything about the overclassification of government documents. This article is just taking current headlines as an excuse to raise a similar sounding but distinct bugbear. There is no indication that any of the documents retained by Trump, Pence, or Biden were overclassified.
The Pence and Biden scandals are most likely the product of using classified information in a busy environment where non-classified info is also used (like the White House). The cliff-edge between administrations where people need to be working with classified information one day, but cleared out of the office the next doesn't help, and the Trump-Biden transition was especially bad.
There is some indication that Trump's motive in keeping the classified documents was for his own ego, therefore his retention of documents again has nothing to do with overclassification, and everything to do with not electing narcissistic assholes to the presidency.
1) Elon's plane flight path is public information because FAA rules require it to be for traffic management purpouses.
2) If Elon wanted, he could by another plane, or $44bn worth of planes, and you'd never know which one he was using.
3) If he flew on commercial planes, or drove, or took the train this would also solve the problem.
4) Elonjet posts where the plane takes off and where it lands. The idea you could use this as practical "assassination coordinates" is ludicrous. You'd need to hover around his plane's last known location, follow it when it takes off, land before him, then rush to get within assassination distance.
So to summarise, his complaint is BS, can easily be solved with money, and he isn't lacking money.
All of your recent articles about Twitter, but this one most of all, make it seem pretty obvious that Elon wants Twitter to be his fiefdom, not a functioning social network site as it has existed in the past.
Elon is driving these changes, often in person, so it seems pretty obvious that he finds the changes desirable.
I absolutely love the fact that 52-48 is the same margin that Brexit won by.
And some slightly more apropos pretentious latin for flavour: Vox populi, ruat caelum
(Ruat caelum roughly translates though the heaven's fall)
Cross-posting all your content is possible, and I assume there is a way of doing so automatically.
But in terms of following and having conversations, it's going to take more time and effort to do so over two platforms, so to my mind it makes sense that you'd designate one your main, even just mentally.
The idea that "Elon Musk @kathygrithin" is impersonating "Elon Musk @elonmusk" is ludicrous. Anyone with 0.5 seconds to spare can look at the permanent, unchangeable account name. This is just a thinly veiled justification for Musk to clamp down on mockery and criticism of himself. I expect more to follow.
Fortunately the excellent @DPRK_News has not been affected yet, but I doubt that will last.
Elon can make the rules now, so he could make it "no publishing any flight tracking" or he could make it "no tracking planes owned by me".
The fact that someone could gather and publish this information elsewhere doesn't negate Elon's control over Twitters rules.
If it had been created as a website it might never have received the attention it did.
This is only valid if you think Elon cares about the quality of the platform and/or the amount of money it makes. He might, he spent a lot of money on it.
But if instead Elon bought Twitter to terminate a few accounts that irritated him (eg the plane tracker), make himself unbanable, and satisfy his ego by being the person who "gave humanity the new public square" and/or "freed twitter from the woke censors", then maybe he's OK with the platform slowly (or quickly) becoming a cesspool.
My guess is that the latter is more true than the former.
The failures I’m talking about are not the failure of the government to step in and kill Kiwi Farms itself — because that would raise serious 1st Amendment issues. It’s the failure to deal with societal issues that make people feel that Kiwi Farms is an appropriate outlet for their own twisted beliefs.
Here's the full paragraph, please read it.
In the second sentence, who has failed to deal with the societal issues, if not the government that was mentioned in the first sentence?
Now explain how I'm being dishonest.
I don't see anything that addresses the terribly inapt analogy (which you specifically agreed with) that cloudflare is like the fire department, a DOS attack is like arson, or that this is about "moral character".
cloudflare is essentially doing the work of private security guards at a convention, ensuring the entrances don't get blocked by protestors. But the protestors can't hurt the attendees, except by blocking access, whereas the attendees are harassing and swatting real people in real life.
A DOS attack against kiwifarms cannot kill anyone, nor permanently destroy property. kiwifarms can return after a DOS attack, once the attackers loose interest or get in legal trouble for performing the attack.
The users, moderators and owners of kiwifarms knowingly and gleefully caused real world suffering and death, and not in the abstract "bacon causes cancer" sense. I don't know what I could say that would be more damning.
It’s the [government] failure to deal with societal issues that make people feel that Kiwi Farms is an appropriate outlet for their own twisted beliefs
So if I've understood you correctly, the government partly to blame for kiwifarms existence because it didn't make a perfect society where everyone is happy.
Remind me, have you ever discussed how easy it is to do content moderation at scale? If so, have you ever considered whether that is applicable to other services performed at whole-of-society scale?
"Some argue that we should terminate these services to content we find reprehensible so that others can launch attacks to knock it offline. That is the equivalent argument in the physical world that the fire department shouldn’t respond to fires in the homes of people who do not possess sufficient moral character."
That is a terrible, utterly inapt analogy. No one is going to die as a result of cloudflare pulling it's services, no physical property will be destroyed, and the digital property can attempt to survive on its own feet.
On top of that, the site wasn't being targeted because its users were not sufficiently virtuous in the eyes of mainstream society, they were engaging in organised doxxing, harassing, and swatting, with the implicit or explicit blessing of the site moderators (if any) and admin.
Here's a much better analogy: cloudflare is like a convention hall. Most of their conventions are normal business, Star Trek, or Anime conventions. They also host a small convention where people discuss how to harass, stalk and swat people, HSSCON. HSSCON participants generate national news coverage, and people start to ask questions about HSSCON and cloudflare for renting the convention centre to HSSCON. Initially cloudflare wants to keep doing business with HSSCON, but a week later concedes.
If HSSCON wants to continue, it still has the option to run the convention in someone's house (analogy for hosting on someone's home computer), or if they have a rich member and/or can pool money, build their own convention centre (which presumably doesn't need to be as big as cloudflare). We've seen this happen before, with Parler, 'Truth' Social etc created as an alternative to Twitter.
"You can still sideload apps onto Android phones, even if they’re not in Google Play (this is in contrast to Apple where things need to go through the app store)."
Or you could use the even simpler method of visiting the website with your phones browser, which would even work on an iDevice.
“cookie popups” on almost every site we visit where we have to click to agree to accept cookies. We know how tiresome these are, and they only require a single click to clear.
If you're only using 1 click then you're accepting all cookies. If you want to reject cookies you always need to make several clicks, especially for sites that mask cookies under "legitimate interest" exceptions.
Firstly the lawsuit alleges that the AI 'read' an illegal copy on the web, not a legally bought copy from a bookshop. Maybe OpenAI will disprove this by pointing to the kindle receipts or whatever, but somehow I doubt that, it's not the techbro way. Secondly, you're implying that it's OK to download the work one word/sentence/page at a time, read it, discard it, then do the same for the next one. We have a word for that, it's called streaming. We know it is legal to stream a movie from netflix/amazon/apple/whoever. But streaming an illegal copy is still illegal, so the underlying source of the material is important. Thirdly I would like a source for the assertion that AI models use streaming to ingest data. It seems to me that by using streaming would be incredibly inefficient. Every time you tweak the model and want to see how it does, you'd need to scan the whole internet again. If you want to control what goes into the model, you need to be able to review the data. Data storage is cheap, and the only thing you need to store is text, which is highly compressible.
Does Twitter have an undelete post function? I'm wondering why they need to keep deleted posts. Even if you need to keep a record that the post happened, you could save space by deleting the contents.
How does "Bungie have plenty of options to deal with cheaters and cheat purveyors" when it can't get the defendant to show up in court, nor collect the judgement against them?
Licencing
I think copyright is the wrong paradigm to look at this case, unless you can prove that they have a copy of your work on their training servers (which I assume they do). The angle to attack this is licencing. Any art that is explicitly posted under a licence with a Non-commercial clause can't be used by a commercial AI model for training, because the company is using it for a commercial project. The CC BY-NC-SA would be an example of that https://creativecommons.org/licenses/by-nc-sa/4.0 Whether the AI generates an exact copy or similar style is irrelevant because commercial usage is a breach of the licence terms. I suspect any art posted without a specific licence can probably do the same because logically the creator retained their rights, including whether it could be used for commercial projects. DeviantArt might have a terms of service argument against this, for art posted on their site, but not art scraped from elsewhere. And OpenAI may have an argument they are non-profit, but wikipedia seems to indicate they aren't just a non-profit anymore.
Yeah, the government under Boris made some clear efforts to influence the actual output of the BBC, and in some cases apparently succeeded. https://www.theguardian.com/media/2023/mar/15/troubling-messages-showing-no-10-pressure-on-bbc-need-investigating-say-former-staff
While the ruling is interesting, the real test is whether it survives appeal. Courts can and do make mistakes, and both the Court of Appeal and Supreme Court of Kenya could overrule this decision. Also because it's issued by a specialised court, I suspect that this ruling would have no weight at other courts within Kenya.
The current scandals do NOT show anything about the overclassification of government documents. This article is just taking current headlines as an excuse to raise a similar sounding but distinct bugbear. There is no indication that any of the documents retained by Trump, Pence, or Biden were overclassified. The Pence and Biden scandals are most likely the product of using classified information in a busy environment where non-classified info is also used (like the White House). The cliff-edge between administrations where people need to be working with classified information one day, but cleared out of the office the next doesn't help, and the Trump-Biden transition was especially bad. There is some indication that Trump's motive in keeping the classified documents was for his own ego, therefore his retention of documents again has nothing to do with overclassification, and everything to do with not electing narcissistic assholes to the presidency.
1) Elon's plane flight path is public information because FAA rules require it to be for traffic management purpouses. 2) If Elon wanted, he could by another plane, or $44bn worth of planes, and you'd never know which one he was using. 3) If he flew on commercial planes, or drove, or took the train this would also solve the problem. 4) Elonjet posts where the plane takes off and where it lands. The idea you could use this as practical "assassination coordinates" is ludicrous. You'd need to hover around his plane's last known location, follow it when it takes off, land before him, then rush to get within assassination distance. So to summarise, his complaint is BS, can easily be solved with money, and he isn't lacking money.
All of your recent articles about Twitter, but this one most of all, make it seem pretty obvious that Elon wants Twitter to be his fiefdom, not a functioning social network site as it has existed in the past. Elon is driving these changes, often in person, so it seems pretty obvious that he finds the changes desirable.
I absolutely love the fact that 52-48 is the same margin that Brexit won by. And some slightly more apropos pretentious latin for flavour: Vox populi, ruat caelum (Ruat caelum roughly translates though the heaven's fall)
Cross-posting all your content is possible, and I assume there is a way of doing so automatically. But in terms of following and having conversations, it's going to take more time and effort to do so over two platforms, so to my mind it makes sense that you'd designate one your main, even just mentally.
The idea that "Elon Musk @kathygrithin" is impersonating "Elon Musk @elonmusk" is ludicrous. Anyone with 0.5 seconds to spare can look at the permanent, unchangeable account name. This is just a thinly veiled justification for Musk to clamp down on mockery and criticism of himself. I expect more to follow. Fortunately the excellent @DPRK_News has not been affected yet, but I doubt that will last.
Elon can make the rules now, so he could make it "no publishing any flight tracking" or he could make it "no tracking planes owned by me". The fact that someone could gather and publish this information elsewhere doesn't negate Elon's control over Twitters rules. If it had been created as a website it might never have received the attention it did.
This is only valid if you think Elon cares about the quality of the platform and/or the amount of money it makes. He might, he spent a lot of money on it. But if instead Elon bought Twitter to terminate a few accounts that irritated him (eg the plane tracker), make himself unbanable, and satisfy his ego by being the person who "gave humanity the new public square" and/or "freed twitter from the woke censors", then maybe he's OK with the platform slowly (or quickly) becoming a cesspool. My guess is that the latter is more true than the former.
That terrible analogy
I don't see anything that addresses the terribly inapt analogy (which you specifically agreed with) that cloudflare is like the fire department, a DOS attack is like arson, or that this is about "moral character".
- cloudflare is essentially doing the work of private security guards at a convention, ensuring the entrances don't get blocked by protestors. But the protestors can't hurt the attendees, except by blocking access, whereas the attendees are harassing and swatting real people in real life.
- A DOS attack against kiwifarms cannot kill anyone, nor permanently destroy property. kiwifarms can return after a DOS attack, once the attackers loose interest or get in legal trouble for performing the attack.
- The users, moderators and owners of kiwifarms knowingly and gleefully caused real world suffering and death, and not in the abstract "bacon causes cancer" sense. I don't know what I could say that would be more damning.
So if I've understood you correctly, the government partly to blame for kiwifarms existence because it didn't make a perfect society where everyone is happy. Remind me, have you ever discussed how easy it is to do content moderation at scale? If so, have you ever considered whether that is applicable to other services performed at whole-of-society scale?inapt analogy
"Some argue that we should terminate these services to content we find reprehensible so that others can launch attacks to knock it offline. That is the equivalent argument in the physical world that the fire department shouldn’t respond to fires in the homes of people who do not possess sufficient moral character." That is a terrible, utterly inapt analogy. No one is going to die as a result of cloudflare pulling it's services, no physical property will be destroyed, and the digital property can attempt to survive on its own feet. On top of that, the site wasn't being targeted because its users were not sufficiently virtuous in the eyes of mainstream society, they were engaging in organised doxxing, harassing, and swatting, with the implicit or explicit blessing of the site moderators (if any) and admin. Here's a much better analogy: cloudflare is like a convention hall. Most of their conventions are normal business, Star Trek, or Anime conventions. They also host a small convention where people discuss how to harass, stalk and swat people, HSSCON. HSSCON participants generate national news coverage, and people start to ask questions about HSSCON and cloudflare for renting the convention centre to HSSCON. Initially cloudflare wants to keep doing business with HSSCON, but a week later concedes. If HSSCON wants to continue, it still has the option to run the convention in someone's house (analogy for hosting on someone's home computer), or if they have a rich member and/or can pool money, build their own convention centre (which presumably doesn't need to be as big as cloudflare). We've seen this happen before, with Parler, 'Truth' Social etc created as an alternative to Twitter.
Actually quite accessible on any platform
"You can still sideload apps onto Android phones, even if they’re not in Google Play (this is in contrast to Apple where things need to go through the app store)." Or you could use the even simpler method of visiting the website with your phones browser, which would even work on an iDevice.