We have a massive poisoning of the commons catastrophe coming, driven by further authoritarian government overreach and control. I've seen no one working on this, and in fact most people on HN seem to be working on ways to further exacerbate this problem. I don't just mean half solutions like tor or social protocols that let you in and out of walled gardens.
There's still a tiny window of opportunity for engineers to come up with or design technical safeguards, but eventually this problem will move past the realm of what's easily solvable and out of our hands, and into policy makers hands. A big part of me feels like that window is already slammed shut.
I agree that it feels like the tiny window of opportunity hasn't quite shut yet, and it's a problem space I know I should take more interest in. What do you see as the viable technical directions? Something along the lines of what Altman was trying to do with his Orb? Something along the lines of the C2PA's Content Credentials?
Instead of collecting biometric info from humans and IDing all of their online movements, you could mandate that LLM output be watermarked (using a technology that Scott Arronson was hired by OpenAI to help develop, after which the project was shut down under Altman right after proving that it could work) so that their online movement would be IDed.
It feels like "Autonomous Coding Agents" are being astroturfed on the daily on HN. The same arguments and tropes are echoing through every thread.
It's hard to distinguish who's a bot, who's a narrative pusher and who's an enthusiast. Which is exactly what you'd want from an astroturfing campaign. There's a clear benefit: people in the industry are reading this, and in doing so they're granting mindshare.
There's one way that can prevent inauthentic support campaigns - personal key signature. But judging by how afraid people, especially in the US, need to be of their government surveilling them, this isn't going to catch on.
>It feels like "Autonomous Coding Agents" are being astroturfed on the daily on HN. The same arguments and tropes are echoing through every thread.
Isn't this what exactly you'd expect in a connected world? The best arguments from both sides proliferate, thereby causing "The same arguments and tropes are echoing through every thread".
> Isn't this what exactly you'd expect in a connected world?
I would expect a figurative war for human attention. With so much information being available, everyone would try to make people focus on what they want to communicate.
> The best arguments
Some of these tropes and arguments aren't really the best. There's a lot of rhetorical gotchas, e.g. "that's exactly what I'd expect from a human" when an automated solution isn't up to par.
> from both sides
The only real "side" is the one actively pushing for something. Everyone else isn't a camp - they're just random people.
Yes. I’ve also been asking every engineer I know what they’re doing with AI and there’s a lot of people doing a lot of different things, but it’s a deep mismatch with the online rhetoric.
This phenomenon appears to be incrementally coming for every single topic and public platform.
I feel the same way. Most people I've talked to are using AI for better search. I don't know anyone using it heavily to do their main job (writing code). I think a lot of the accounts bragging about how much they are doing with AI are bots.
I'm even shocked when I hear people are using it for better search. I've found it to be terrible for search, and constantly fabricating things. It's distilled everything that is bad about new Google, where it prefers popular results to accurate ones - but with actual fabrication that becomes infinitely worse.
I literally ask it to look for something, and immediately afterwards (before reading the long-winded result), ask it if the results were real or fabricated. It's just how the cost-benefit analysis works out, and I didn't learn until a ton of times reading the results, getting suspicious of a few, doing websearches to verify them, not finding them, then coming back to ask if they were real.
"Sorry! It's absolutely fair that you called me out on that... It's important that you hold me to a high standard... You're absolutely right..."
I'm finding it valuable for compressing all of the docs in the world, so I don't have to look up what a function does or how to accomplish something in some framework or CLI. I find it capable of writing code if I move an inch at a time; build copious verbose debugging output that I feed back into it every time it screws up; and when it goes into a stupid loop being stupid, just debugging by hand before wasting hours trying to get it to see something that it doesn't want to see.
It feels the same way on GitHub trending. I used to check it frequently to see what the hottest newest tech was and stay up to date. Now it's oversaturated by whatever the newest AI bubble is. It also doesn't help that MCP enabled products like OpenClaw star their own repo and artificially inflate their perceived value.
I hate to sound like I’m turfing for cryptocurrencies, isn’t there like an identity solution there that the crypto nerds solved for to keep identity verification anonymous and surveillance proof?
Need to double check what is available, though I feel like that angle could work.
I’ve been wondering also if a simple lie & deception detection type system could be a useful angles. It’s complicated in practice; though the human intuition would say it’s figured this out millennia ago- I can’t tell you how many times my body has figured out someone’s toxic negative vibe by feeling. And I think we probably understand this better than we think and can represent it in the computer space with analysis of signals and some follow on questions. Hope I’m not too naive here.
If you can point me at someone that would fund such projects (not VCs), would be happy to apply. Projects like NLNet aren't keen on funding larger scope projects. At least if you do not have the thought leader influencer clout.
Discussion of astroturfing on a post that is specifically about astroturfing is such an obvious exception that I'm having a hard time taking your reply in good faith, but this is me trying to do so anyway instead of just downvoting and flagging like the guidelines suggest I do in such cases.
To quote The Cable Guy, there’s only one answer, someone has to kill the babysitter (tv, social media, Big Tech). It’s hard to kill the babysitter when everyone in Congress is invested balls deep in the babysitter. Eisenhower warned of the coming overreaching powers of the Military Industrial Complex, but no one is attacking the Government Stock Market Tech Complex (GSMTC).
It’s beyond that. It’s the CIA deeply embedded in all the scary uncomfortable ways you would have hoped never possible. Presidents win and turn their stance and run around in the other direction, they don’t what to be another assassinated Kennedy (and imo today they would have other fears worse than dying). Congressmen and women are definitely also aware of the deep presence and power of that agency and its perversion into American life and politics. They don’t want to be the ones to be the sacrificial pawn sparking an outright violent American revolution and tear down of the agency.
I was surveilled, experimented on and followed by them for being American-Pakistani and speaking out against them from 2022-2023. It was a scary time and I wish I were making this up. I wonder sometimes if they really are the good guys, and I just got things backwards. I also heard when you are kidnapped and in hostile territories for long enough, you fall in love with the kidnappers.
There were many disinformation research organizations in the US, including at major institutions such as Harvard and Stanford, that were forced to close by conservatives through lawfare or apparently through donor pressure.
(It's interesting that conservatives saw it as a partisan cause.)
their landing page stops short of saying that Doublespeed would be "a good fit for your political campaign." I'd prefer fighting an AI-powered drone over becoming a victim of "Dead Internet-aaS" startup. at least, flying lawnmowers are honest
strong agree, I feel like it poisons the fabric of society somehow when everything you interact with is fake or even just has a good chance of being fake, regardless of the also-shitty fact that it is also often trying to influence you.
Also how the being fake doesn't even have to be malicious. now every tom, dick, and harry wants to create content. All the world's a stage, follower count go up.
I held a hope that it would create an evolutionary pressure that would weed out people who fall for foolish arguments i.e. arguments without any sort of structure that should be capable of convincing anyone of anything. But that's just wishful thinking. People fall for anything as long as it's flattering and it allows them to do what they want to do when they want to do it.
Every propagandistic argument is going to be like that for 80% of people, and 40% of people are going to be within that 80% about 99% of the time. They think the biggest issue of our time is how much people complain.
My browser highlights a few hundred accounts. For HN and other comment-oriented sites, local userscripts are supported by browser plugins, including mobile Safari. These can highlight known usernames and implement blocklists. Most LLMs can generate a userscript on demand for non-obfuscated sites, including userid list for manual edit.
This is notorious in platforms like reddit, with people jumping in to suggest no name products in response to questions. It doesn't help that reddit allows private profiles, thus allowing astroturfers to get away with it. Also, another case is LLM astroturfing, we're bombarded with doomerism and obituaries about programming, some of said opinions are subtler, short comments, the most dangerous ones, because little by little they jab you, though the most conspicuous ones are easy to identify. And then there's the political astroturfing. In my country smokescreens are the defacto tool, but it is suspicious of the amount of high quality edits and memes that came out about the Epstein files, essentially cementing him as a "meme" and not a monster that abused minors.
We have a massive poisoning of the commons catastrophe coming, driven by further authoritarian government overreach and control. I've seen no one working on this, and in fact most people on HN seem to be working on ways to further exacerbate this problem. I don't just mean half solutions like tor or social protocols that let you in and out of walled gardens.
There's still a tiny window of opportunity for engineers to come up with or design technical safeguards, but eventually this problem will move past the realm of what's easily solvable and out of our hands, and into policy makers hands. A big part of me feels like that window is already slammed shut.
I agree that it feels like the tiny window of opportunity hasn't quite shut yet, and it's a problem space I know I should take more interest in. What do you see as the viable technical directions? Something along the lines of what Altman was trying to do with his Orb? Something along the lines of the C2PA's Content Credentials?
[0] e.g. https://www.businessinsider.com/sam-altman-tools-for-humanit... and the feature piece at https://time.com/7288387/sam-altman-orb-tools-for-humanity/
[1] https://contentcredentials.org and https://c2pa.org
Instead of collecting biometric info from humans and IDing all of their online movements, you could mandate that LLM output be watermarked (using a technology that Scott Arronson was hired by OpenAI to help develop, after which the project was shut down under Altman right after proving that it could work) so that their online movement would be IDed.
It feels like "Autonomous Coding Agents" are being astroturfed on the daily on HN. The same arguments and tropes are echoing through every thread.
It's hard to distinguish who's a bot, who's a narrative pusher and who's an enthusiast. Which is exactly what you'd want from an astroturfing campaign. There's a clear benefit: people in the industry are reading this, and in doing so they're granting mindshare.
There's one way that can prevent inauthentic support campaigns - personal key signature. But judging by how afraid people, especially in the US, need to be of their government surveilling them, this isn't going to catch on.
>It feels like "Autonomous Coding Agents" are being astroturfed on the daily on HN. The same arguments and tropes are echoing through every thread.
Isn't this what exactly you'd expect in a connected world? The best arguments from both sides proliferate, thereby causing "The same arguments and tropes are echoing through every thread".
> Isn't this what exactly you'd expect in a connected world?
I would expect a figurative war for human attention. With so much information being available, everyone would try to make people focus on what they want to communicate.
> The best arguments
Some of these tropes and arguments aren't really the best. There's a lot of rhetorical gotchas, e.g. "that's exactly what I'd expect from a human" when an automated solution isn't up to par.
> from both sides
The only real "side" is the one actively pushing for something. Everyone else isn't a camp - they're just random people.
Yes. I’ve also been asking every engineer I know what they’re doing with AI and there’s a lot of people doing a lot of different things, but it’s a deep mismatch with the online rhetoric.
This phenomenon appears to be incrementally coming for every single topic and public platform.
I feel the same way. Most people I've talked to are using AI for better search. I don't know anyone using it heavily to do their main job (writing code). I think a lot of the accounts bragging about how much they are doing with AI are bots.
I'm even shocked when I hear people are using it for better search. I've found it to be terrible for search, and constantly fabricating things. It's distilled everything that is bad about new Google, where it prefers popular results to accurate ones - but with actual fabrication that becomes infinitely worse.
I literally ask it to look for something, and immediately afterwards (before reading the long-winded result), ask it if the results were real or fabricated. It's just how the cost-benefit analysis works out, and I didn't learn until a ton of times reading the results, getting suspicious of a few, doing websearches to verify them, not finding them, then coming back to ask if they were real.
"Sorry! It's absolutely fair that you called me out on that... It's important that you hold me to a high standard... You're absolutely right..."
I'm finding it valuable for compressing all of the docs in the world, so I don't have to look up what a function does or how to accomplish something in some framework or CLI. I find it capable of writing code if I move an inch at a time; build copious verbose debugging output that I feed back into it every time it screws up; and when it goes into a stupid loop being stupid, just debugging by hand before wasting hours trying to get it to see something that it doesn't want to see.
It feels the same way on GitHub trending. I used to check it frequently to see what the hottest newest tech was and stay up to date. Now it's oversaturated by whatever the newest AI bubble is. It also doesn't help that MCP enabled products like OpenClaw star their own repo and artificially inflate their perceived value.
Interesting - claw faking the benchmarks .. they match well with openA ideologically .
I hate to sound like I’m turfing for cryptocurrencies, isn’t there like an identity solution there that the crypto nerds solved for to keep identity verification anonymous and surveillance proof?
Need to double check what is available, though I feel like that angle could work.
I’ve been wondering also if a simple lie & deception detection type system could be a useful angles. It’s complicated in practice; though the human intuition would say it’s figured this out millennia ago- I can’t tell you how many times my body has figured out someone’s toxic negative vibe by feeling. And I think we probably understand this better than we think and can represent it in the computer space with analysis of signals and some follow on questions. Hope I’m not too naive here.
If you can point me at someone that would fund such projects (not VCs), would be happy to apply. Projects like NLNet aren't keen on funding larger scope projects. At least if you do not have the thought leader influencer clout.
What are your ideas for this?
> I've seen no one working on this, and in fact most people on HN seem to be working on ways to further exacerbate this problem.
It's against the HN guidelines to insinuate that astroturfing happens on HN.
Discussion of astroturfing on a post that is specifically about astroturfing is such an obvious exception that I'm having a hard time taking your reply in good faith, but this is me trying to do so anyway instead of just downvoting and flagging like the guidelines suggest I do in such cases.
To quote The Cable Guy, there’s only one answer, someone has to kill the babysitter (tv, social media, Big Tech). It’s hard to kill the babysitter when everyone in Congress is invested balls deep in the babysitter. Eisenhower warned of the coming overreaching powers of the Military Industrial Complex, but no one is attacking the Government Stock Market Tech Complex (GSMTC).
It’s beyond that. It’s the CIA deeply embedded in all the scary uncomfortable ways you would have hoped never possible. Presidents win and turn their stance and run around in the other direction, they don’t what to be another assassinated Kennedy (and imo today they would have other fears worse than dying). Congressmen and women are definitely also aware of the deep presence and power of that agency and its perversion into American life and politics. They don’t want to be the ones to be the sacrificial pawn sparking an outright violent American revolution and tear down of the agency.
I was surveilled, experimented on and followed by them for being American-Pakistani and speaking out against them from 2022-2023. It was a scary time and I wish I were making this up. I wonder sometimes if they really are the good guys, and I just got things backwards. I also heard when you are kidnapped and in hostile territories for long enough, you fall in love with the kidnappers.
Happy to share more details if anyone’s curious.
so what you're saying is that the US government is an illegitimate regime and everyone can fully justify destroying it as an enemy of the people?
The CIA is not the US government.
Its already here.
There were many disinformation research organizations in the US, including at major institutions such as Harvard and Stanford, that were forced to close by conservatives through lawfare or apparently through donor pressure.
(It's interesting that conservatives saw it as a partisan cause.)
related: https://doublespeed.ai/ - basically astroturfing as a service.
their landing page stops short of saying that Doublespeed would be "a good fit for your political campaign." I'd prefer fighting an AI-powered drone over becoming a victim of "Dead Internet-aaS" startup. at least, flying lawnmowers are honest
strong agree, I feel like it poisons the fabric of society somehow when everything you interact with is fake or even just has a good chance of being fake, regardless of the also-shitty fact that it is also often trying to influence you.
Also how the being fake doesn't even have to be malicious. now every tom, dick, and harry wants to create content. All the world's a stage, follower count go up.
I held a hope that it would create an evolutionary pressure that would weed out people who fall for foolish arguments i.e. arguments without any sort of structure that should be capable of convincing anyone of anything. But that's just wishful thinking. People fall for anything as long as it's flattering and it allows them to do what they want to do when they want to do it.
Every propagandistic argument is going to be like that for 80% of people, and 40% of people are going to be within that 80% about 99% of the time. They think the biggest issue of our time is how much people complain.
My browser highlights a few hundred accounts. For HN and other comment-oriented sites, local userscripts are supported by browser plugins, including mobile Safari. These can highlight known usernames and implement blocklists. Most LLMs can generate a userscript on demand for non-obfuscated sites, including userid list for manual edit.
This is notorious in platforms like reddit, with people jumping in to suggest no name products in response to questions. It doesn't help that reddit allows private profiles, thus allowing astroturfers to get away with it. Also, another case is LLM astroturfing, we're bombarded with doomerism and obituaries about programming, some of said opinions are subtler, short comments, the most dangerous ones, because little by little they jab you, though the most conspicuous ones are easy to identify. And then there's the political astroturfing. In my country smokescreens are the defacto tool, but it is suspicious of the amount of high quality edits and memes that came out about the Epstein files, essentially cementing him as a "meme" and not a monster that abused minors.
(2022)