Section 230 would have been fine if narrowly, instead of maximally, interpreted.
There is always going to be some offense jockey looking for a way to peddle feigned outrage into a tidy profit. Without Section 230 we would be flooded by endless frivolous lawsuits over everything ever posted anonymously in a comment section, as I am doing right now. Most of these would fail on merits, but the financial burden of defending against them would make it nearly impossible to profitably run internet infrastructure at any level.
I'm open to a better solution to the same problem. And certainly it shouldn't shield service providers from their own content and editorial choices, whether human or algorithmic. That was an unfortunate interpretation of the law from the start. But I don't think we can do away with it unless we come up with a viable replacement in the process. Or, can burn the internet down. That would be fine I guess.
Repealing Section 230 wouldn't burn down the internet; it would just kill the corporate surveillance state. Without this liability shield, the web would naturally shift toward decentralized P2P networks free of ads, tracking, bots, and algorithmic manipulation and rage-bait—much like PeerTube operates today.
The collapse of Section 230 would be a problem for tech shareholders, sure, but it would be an absolute boon for anyone who values an honest exchange of ideas on the internet.
Did you see that bit in the quote where Senator Durbin says he can't remember anyone coming to Congress and asking to be regulated before?
The reason that the big platforms want to get rid of Section 230 is that they ARE big enough to moderate everything, and they realize that would *give them a moat* against new competitors, who would be unable to start a competing social media platform without enough resources to moderate everything. It would mandate that every platform MUST maintain a corporate surveillance state.
That definitely makes sense, and I can see how you came to that conclusion. And I don't want to give the impression that repealing Section 230 would magically remove the corporate surveillance state. It won't.
However, if we make some assumptions, I can conclude that repealing Section 230 won't give LLM corps the moat they think they would get, and might possibly even be a multi-billion-dollar hassle.
Assumption 1) Big platforms, specifically LLM platforms, that are big enough to moderate everything do so through either labor costs or software costs. In other words, they hire human moderators or task LLM moderation systems.
Assumption 2) LLMs will continue to hallucinate, and for important tasks (e.g., compliance-based tasks), humans will need to give final approval. This is why LLMs, like ChatGPT, still display a disclaimer along with any generated text, e.g., "ChatGPT can make mistakes. Check important info."
Assumption 3) Though less accurate, LLM-based content moderation is cheaper and faster than human-based content moderation.
Assumption 4) Big companies are unconcerned with less accurate LLM-based content moderation because Section 230 shields them from false negatives.
With these assumptions in place, I can see how their moat is smaller than they would hope, because these assumptions seem to necessitate the need to hire expensive, human-based content moderation if Section 230 is repealed. Not saying their moat disappears when that happens, but it's not the powerplay it first seems to be if these assumptions hold. I think what would happen instead is that big platforms would scale back user-generated content.
I'm not so confident that such a renaissance of users opening their minds about P2P would occur, or whether it would be a good thing if they did.
On the former, IRL normal people don't understand, care or want to know why P2P is “quirky” and unreliable. To them it's just bad UX. Same goes with other nerdtech. Concrete example, LBRY is amazing P2P streaming tech. Nobody wants to download the app and watch it drain batteries/resources being a good peer, nor do they want to hear excuses for why they click a video and it doesn't instantly start streaming. They just say it sucks and walk away. Can't even get them to use the normie frontend, odyssee. Because the videos are slow and unreliable and the search is weird and there's not this and that feature they're used to from Youtube/twitch. They don't care that if they follow an easy 27 step process they can kinda sorta emulate that.
On the latter, it's not that I like surveillance capitalism or want people to be subjected to it. But there are only a few ways to handle an Eternal September flood on a decentralized network, and one looks like 4chan with its total shitposting chaos, the other looks like fediverse with its endless petty tyrannies and volunteer stasi brigades. Neither are particularly appealing.
Then there’s the other problem of: as soon as pols and litigious types have the power to exercise authority over content on the mainstream web, the very next thing they'll want is to make sure you can't escape it. They'll go after P2P immediately, if nothing else by prosecuting ISPs for routing it, who will no longer have section 230 shields. And contrary to belligerent hacker hopes, they can effectively inhibit it with existing technology. It's just not very cost effective when they can't be sued for refusing.
See also the absolute nightmare that is the current draft IPV8 spec, which if implemented will kill privacy and autonomy dead for anybody who isn't a full-time superhacker. We unhappy few will be building our own hardware layer at that point out of tin cans and strings, and ordinary people will not be involved.
Anyway, sorry to rain on your parade. It's just, we've been talking mad shit about what we'd do in these kinds of scenarios for decades, but IRL things don't work that way. It's a fun LARP, but if the internet as we know it really does burn down I think I'll just spend even more time on the beach and in my garden instead.
Frictionless corporate platforms just prop up surveillance-advertisement, at best. If casual users find P2P too clunky and choose to walk away, they truly aren't missing out on yet another on-trend, bizarre video to watch while pooping. Everyone's FOMO will eventually subside.
Further, in my experience, fractured, self-governing niche spaces are just how human communities naturally operate offline. Section 230 doesn't fix that.
So, sounds like this is a case of "one person's trash is another's treasure".
If killing Section 230 leads to people logging off to spend more time on the beach and in their gardens, then feel free to send me an invite. Sounds like a great time!
Exactly why I'm unironically fine with burning it down. People spend too much time on computers and it's making them miserable. Most of what we hoped the internet would be didn't work out. Corponet covers some of the remainder, whatever weird hacker remnant thing will cover the rest. But for the most part breaking it can only be good in the longrun. After 30 years of watching the dream turn into a nightmare, I think it's time to wake up.
Agreed, but what we're seeing is the general state of litigiousness that pervades the U.S. This means a law has to be written in such specific terms that it becomes difficult to apply generally, or that everyone has to be prepared for the lawyers to smell the blood and come looking for the feast...
Section 230 is one the most misunderstood and misrepresented laws. It has never been about blanket liability for platforms. It only applies to 3rd party (user) content posted to the platform. It allows platforms to moderate content as it is reported, which takes time. Moderating at scale is impossible. A platform simply cannot knowingly ignore user criminal/illegal content and use section 230 as a shield, that’s not how it works.
Someone has to be accountable for all this text, and Gemini and other chatbots do extrude text that isn’t in the sources.
Another one, Gary. Taylor Lorenz got Brockman to admit funding sock puppet doomer and anti doomer social media accounts. Their commitment to anti truth is undefeated.
This isn’t actually the first ruling like this. A couple of years back a Canadian court ruled that Air Canada was liable for promises made by the LLM chatbot that was acting as a customer service agent on Air Canada’s website.
Thanks for the reminder. It seemed amazing at the time that the question even had to be litigated. Of course we should be liable for what our agents say, human and otherwise.
It seems to me, though, that you can make an argument that what emanates from ChatGPT is different, in that it's not in direct control. I think you might instead find yourself arguing that there's a defect in the product. It'll be interesting to read the German decision.
And, oh, by the way, Section 230 wasn't originally about giving an industry time to establish itself. It was about protecting internet service providers from liability for what traveled over their facilities, analogous to what had always been the case for common carriers like the telephone network.
As soon as Facebook and the rest of the industry started pushing content at users, they became publishers, and their claim of immunity from liability ought to have vanished into thin air. An algorithm can't be held responsible for its editorial decisions, but the company that implemented it surely ought to have been. It was sheer technical ignorance on the part of the court system that this didn't happen. Perhaps not too late to start now?
And I thought the day couldn't get much better. No apologies required for this one either. Hope for accountability and a better future, at last! I've been writing my reps for years about the horrors of Section 230, but never even got a reply other than vague, unrelated platitudes.
That Illinois law is a slap in the face, providing zip for damages to the masses while continuing to protect the frauds and billionaires from the damages they cause—all while masquerading as regulation.
And another thing, US libertarians, please stop saying the nonsense that in USA you don't regulate as much as in EU. You are regulating that those companies can not be sued. You are regulating that nobody can make them responsible of their own products. That is the strongest possible regulation. You are putting those companies over any law other companies have to follow. They could sell poison and you will regulate in favor. Free market? Don't make me laugh.
Working class engineer repeating that probability of error in LLMs follow a power law here. Nobody is going to insurance a sigma event generator machine. No-f***-body. This is why I have already seen 2 startups in Y-combinator with insurance for agents as business model. Well, the real business model is to have legal and funds in Bahamas or Bermuda, and run with the money as soon as the big claim appears (it will appear 100% sure about that, the law of the big numbers and all that stuff). They know it. Nobody is going to insurance a sigma event generator machine. So what regulation can not do, the market will destroy. Germany is the first pointing, but EU will follow. GDPR is a beautiful acronym, you'll see people speaking about that also.
Well Scam may be for regulation now if it keeps his competitors from getting too far ahead of him.To paraphrase Karl Bodde, the tech isnt the problem, it is all the people behind it that we need to worry about. Talk about terrible people -and terrible liars, all of them. They just care about how much money it can make them. Screw everyone else
I'm particularly amused that the only likely defense for Section 230 protection of AI agents is - that they were all trained by learning (aka stealing) human created documents. They are caught between claiming their products are truly "intelligent" and face legal action, or admitting they are simply "regurgitation" which invalidates their claimed abilities and value proposition.
Just to remind that in its recent Cox Communications decision, the Supreme Court actually expanded the freedom from liability afforded by Section 230. Copyright infringement is excluded from Section 230 immunity. But in Cox, the Court may have effectively closed the loophole in that immunity by requiring an author or other copyright owner to prove intent on the part of the ISP. The US courts won't help, so I think any remedy to the 230 problem lies, if anywhere, in Congress. That said, the odds are low as mentioned.
Section 230 would have been fine if narrowly, instead of maximally, interpreted.
There is always going to be some offense jockey looking for a way to peddle feigned outrage into a tidy profit. Without Section 230 we would be flooded by endless frivolous lawsuits over everything ever posted anonymously in a comment section, as I am doing right now. Most of these would fail on merits, but the financial burden of defending against them would make it nearly impossible to profitably run internet infrastructure at any level.
I'm open to a better solution to the same problem. And certainly it shouldn't shield service providers from their own content and editorial choices, whether human or algorithmic. That was an unfortunate interpretation of the law from the start. But I don't think we can do away with it unless we come up with a viable replacement in the process. Or, can burn the internet down. That would be fine I guess.
Repealing Section 230 wouldn't burn down the internet; it would just kill the corporate surveillance state. Without this liability shield, the web would naturally shift toward decentralized P2P networks free of ads, tracking, bots, and algorithmic manipulation and rage-bait—much like PeerTube operates today.
The collapse of Section 230 would be a problem for tech shareholders, sure, but it would be an absolute boon for anyone who values an honest exchange of ideas on the internet.
Did you see that bit in the quote where Senator Durbin says he can't remember anyone coming to Congress and asking to be regulated before?
The reason that the big platforms want to get rid of Section 230 is that they ARE big enough to moderate everything, and they realize that would *give them a moat* against new competitors, who would be unable to start a competing social media platform without enough resources to moderate everything. It would mandate that every platform MUST maintain a corporate surveillance state.
That definitely makes sense, and I can see how you came to that conclusion. And I don't want to give the impression that repealing Section 230 would magically remove the corporate surveillance state. It won't.
However, if we make some assumptions, I can conclude that repealing Section 230 won't give LLM corps the moat they think they would get, and might possibly even be a multi-billion-dollar hassle.
Assumption 1) Big platforms, specifically LLM platforms, that are big enough to moderate everything do so through either labor costs or software costs. In other words, they hire human moderators or task LLM moderation systems.
Assumption 2) LLMs will continue to hallucinate, and for important tasks (e.g., compliance-based tasks), humans will need to give final approval. This is why LLMs, like ChatGPT, still display a disclaimer along with any generated text, e.g., "ChatGPT can make mistakes. Check important info."
Assumption 3) Though less accurate, LLM-based content moderation is cheaper and faster than human-based content moderation.
Assumption 4) Big companies are unconcerned with less accurate LLM-based content moderation because Section 230 shields them from false negatives.
With these assumptions in place, I can see how their moat is smaller than they would hope, because these assumptions seem to necessitate the need to hire expensive, human-based content moderation if Section 230 is repealed. Not saying their moat disappears when that happens, but it's not the powerplay it first seems to be if these assumptions hold. I think what would happen instead is that big platforms would scale back user-generated content.
I'm not so confident that such a renaissance of users opening their minds about P2P would occur, or whether it would be a good thing if they did.
On the former, IRL normal people don't understand, care or want to know why P2P is “quirky” and unreliable. To them it's just bad UX. Same goes with other nerdtech. Concrete example, LBRY is amazing P2P streaming tech. Nobody wants to download the app and watch it drain batteries/resources being a good peer, nor do they want to hear excuses for why they click a video and it doesn't instantly start streaming. They just say it sucks and walk away. Can't even get them to use the normie frontend, odyssee. Because the videos are slow and unreliable and the search is weird and there's not this and that feature they're used to from Youtube/twitch. They don't care that if they follow an easy 27 step process they can kinda sorta emulate that.
On the latter, it's not that I like surveillance capitalism or want people to be subjected to it. But there are only a few ways to handle an Eternal September flood on a decentralized network, and one looks like 4chan with its total shitposting chaos, the other looks like fediverse with its endless petty tyrannies and volunteer stasi brigades. Neither are particularly appealing.
Then there’s the other problem of: as soon as pols and litigious types have the power to exercise authority over content on the mainstream web, the very next thing they'll want is to make sure you can't escape it. They'll go after P2P immediately, if nothing else by prosecuting ISPs for routing it, who will no longer have section 230 shields. And contrary to belligerent hacker hopes, they can effectively inhibit it with existing technology. It's just not very cost effective when they can't be sued for refusing.
See also the absolute nightmare that is the current draft IPV8 spec, which if implemented will kill privacy and autonomy dead for anybody who isn't a full-time superhacker. We unhappy few will be building our own hardware layer at that point out of tin cans and strings, and ordinary people will not be involved.
Anyway, sorry to rain on your parade. It's just, we've been talking mad shit about what we'd do in these kinds of scenarios for decades, but IRL things don't work that way. It's a fun LARP, but if the internet as we know it really does burn down I think I'll just spend even more time on the beach and in my garden instead.
Frictionless corporate platforms just prop up surveillance-advertisement, at best. If casual users find P2P too clunky and choose to walk away, they truly aren't missing out on yet another on-trend, bizarre video to watch while pooping. Everyone's FOMO will eventually subside.
Further, in my experience, fractured, self-governing niche spaces are just how human communities naturally operate offline. Section 230 doesn't fix that.
So, sounds like this is a case of "one person's trash is another's treasure".
If killing Section 230 leads to people logging off to spend more time on the beach and in their gardens, then feel free to send me an invite. Sounds like a great time!
Exactly why I'm unironically fine with burning it down. People spend too much time on computers and it's making them miserable. Most of what we hoped the internet would be didn't work out. Corponet covers some of the remainder, whatever weird hacker remnant thing will cover the rest. But for the most part breaking it can only be good in the longrun. After 30 years of watching the dream turn into a nightmare, I think it's time to wake up.
Agreed, but what we're seeing is the general state of litigiousness that pervades the U.S. This means a law has to be written in such specific terms that it becomes difficult to apply generally, or that everyone has to be prepared for the lawyers to smell the blood and come looking for the feast...
Section 230 is one the most misunderstood and misrepresented laws. It has never been about blanket liability for platforms. It only applies to 3rd party (user) content posted to the platform. It allows platforms to moderate content as it is reported, which takes time. Moderating at scale is impossible. A platform simply cannot knowingly ignore user criminal/illegal content and use section 230 as a shield, that’s not how it works.
You are right, the German ruling is profound. I hope this points the way to meaningful regulation.
Someone has to be accountable for all this text, and Gemini and other chatbots do extrude text that isn’t in the sources.
Another one, Gary. Taylor Lorenz got Brockman to admit funding sock puppet doomer and anti doomer social media accounts. Their commitment to anti truth is undefeated.
Some of them are also very committed to Miss Auntie Thropic.
“Miss Aunt Thropic”
(“To Serve Man” — hat tip to Rod Serling)
Miss Aunt Thropic
Serves the human
Every topic
Turns to cumin
Other spice
Will also serve
Auntie’s vice
For man hor d'oeuvre
This isn’t actually the first ruling like this. A couple of years back a Canadian court ruled that Air Canada was liable for promises made by the LLM chatbot that was acting as a customer service agent on Air Canada’s website.
Thanks for the reminder. It seemed amazing at the time that the question even had to be litigated. Of course we should be liable for what our agents say, human and otherwise.
It seems to me, though, that you can make an argument that what emanates from ChatGPT is different, in that it's not in direct control. I think you might instead find yourself arguing that there's a defect in the product. It'll be interesting to read the German decision.
And, oh, by the way, Section 230 wasn't originally about giving an industry time to establish itself. It was about protecting internet service providers from liability for what traveled over their facilities, analogous to what had always been the case for common carriers like the telephone network.
As soon as Facebook and the rest of the industry started pushing content at users, they became publishers, and their claim of immunity from liability ought to have vanished into thin air. An algorithm can't be held responsible for its editorial decisions, but the company that implemented it surely ought to have been. It was sheer technical ignorance on the part of the court system that this didn't happen. Perhaps not too late to start now?
Why are you still on X?
And I thought the day couldn't get much better. No apologies required for this one either. Hope for accountability and a better future, at last! I've been writing my reps for years about the horrors of Section 230, but never even got a reply other than vague, unrelated platitudes.
That Illinois law is a slap in the face, providing zip for damages to the masses while continuing to protect the frauds and billionaires from the damages they cause—all while masquerading as regulation.
“never even got a reply other than vague, unrelated platitudes.“
At least years ago you used to get platitudes written by a human.
Now you just get vague, unrelated botitudes (attitudes expressed by bots) if you get a reply at all
They were equally bad, though. Lately each paragraph is in a different sized font; too hard for the LLMs to figure that out?
That’s to make it look like it was not generated by a bot.
Like adding spelling errors and beginning sentences with lower case and other ruses.
And another thing, US libertarians, please stop saying the nonsense that in USA you don't regulate as much as in EU. You are regulating that those companies can not be sued. You are regulating that nobody can make them responsible of their own products. That is the strongest possible regulation. You are putting those companies over any law other companies have to follow. They could sell poison and you will regulate in favor. Free market? Don't make me laugh.
Working class engineer repeating that probability of error in LLMs follow a power law here. Nobody is going to insurance a sigma event generator machine. No-f***-body. This is why I have already seen 2 startups in Y-combinator with insurance for agents as business model. Well, the real business model is to have legal and funds in Bahamas or Bermuda, and run with the money as soon as the big claim appears (it will appear 100% sure about that, the law of the big numbers and all that stuff). They know it. Nobody is going to insurance a sigma event generator machine. So what regulation can not do, the market will destroy. Germany is the first pointing, but EU will follow. GDPR is a beautiful acronym, you'll see people speaking about that also.
Well Scam may be for regulation now if it keeps his competitors from getting too far ahead of him.To paraphrase Karl Bodde, the tech isnt the problem, it is all the people behind it that we need to worry about. Talk about terrible people -and terrible liars, all of them. They just care about how much money it can make them. Screw everyone else
that was 2023
How quickly the time goes by when one is having fun.
I'm particularly amused that the only likely defense for Section 230 protection of AI agents is - that they were all trained by learning (aka stealing) human created documents. They are caught between claiming their products are truly "intelligent" and face legal action, or admitting they are simply "regurgitation" which invalidates their claimed abilities and value proposition.
https://substack.com/profile/396786532-arun-varghese/note/c-274846111?r=6k8ig4&utm_source=notes-share-action&utm_medium=web
Just to remind that in its recent Cox Communications decision, the Supreme Court actually expanded the freedom from liability afforded by Section 230. Copyright infringement is excluded from Section 230 immunity. But in Cox, the Court may have effectively closed the loophole in that immunity by requiring an author or other copyright owner to prove intent on the part of the ISP. The US courts won't help, so I think any remedy to the 230 problem lies, if anywhere, in Congress. That said, the odds are low as mentioned.
“Lie-ability”
Lie-ability
Plagues AI’s
Same agility
Pays their guys
Time for the law suits to begin.