Hacker News

points by slg 3 days ago | 0 comments

I don't think people are ready for the amount of work AI agents are going to be creating for real humans. We're effectively putting the tools of mass spammers in the hands of everyone and it's going to be incredibly frustrating for all involved. "If it takes me 30 seconds to fire off a prompt that has potential to benefit me it's worth it regardless of the 2 minutes of work it creates for a dozen people" scaled to all of society.

[–] Aurornis 3 days ago

> "If it takes me 30 seconds to fire off a prompt that has potential to benefit me it's worth it regardless of the 2 minutes of work it creates for a dozen people" scaled to all of society.

People love this stuff if it benefits them. It even gets cheered on when the receiving end can be reduced to something disliked, like a big corporation or a bureaucracy.

There was a startup that would fight traffic and parking tickets for you and everyone loved the idea. Until, that is, you realize that the company is doing automated requests to jam up your local government systems with extra work. The ticketing agents and government workers who have to put in extra time on these cases don't really care because it's no different than any other work. In some cases they might be able to get some overtime.

The real losers are people who have legitimate complaints who get backlogged behind a thousand automated requests from people who asked ChatGPT to try to fight something for them.

Then later, we all lose when the only way to contest things is to go to a physical location in person and prove that it's you trying to do it, not the bot.

Already seeing this with job applications, too: Apply for a job and you'll be mixed in with 1000 automated submissions. Do the interview remotely and they'll be wondering if you're cheating. Now you have to fly on site for middle interview stages to prove that it's you, not a bot, who can answer the questions.

[–] hansonkd 3 days ago

Thats assuming the people using automated tools don't have legitimate complaints and putting the blame on them for their government's failure. If the system was designed better and responded better, maybe less people would have to turn to spamming the IRS line (as an example) when it opens for a glimmer of hope of talking to a person.

Same with the job search. if so many companies didn't make it a pain to input your information, maybe they could get more sympathy.

[–] Aurornis 3 days ago

Nobody thinks that everyone using automated services to beat parking tickets or whatever was actually being wrongfully accused. Most just want to find a loophole to get out of it.

You're missing the important part where people with legitimate complaints are being crowded out by these frivolous attempts to avoid fines. Everyone loses when filing automated frivolous complaints becomes the default response to getting a parking ticket, for example.

> Same with the job search. if so many companies didn't make it a pain to input your information, maybe they could get more sympathy.

I think you’re confused about the problem. I wasn’t asking you to be sympathetic to employers. I’m saying that the application and interview system abusers are ruining it for real, honest job applicants.

Again, the spammers are taking away from legitimate applicants

[–] jimmydorry 3 days ago

Why is the default assumption that these attempts to contest parking fines are frivolous? Have you never received a fine for parking legally? The fines are created by a human after all. If the system was perfect, they wouldn't offer a way to contest them.

[–] nhinck2 3 days ago

Because I've talked to other human beings before?

[–] Aurornis 2 days ago

Because that’s how the services were advertised? And that’s how people operate?

If you create a service that might get someone out of a fine via a loophole or by creating friction, many will take that opportunity to take the chance.

It would be very unreasonable to think that most of the people trying these services were simply wronged by the government and we must support them all.

> Have you never received a fine for parking legally?

I have not, but I have had to deal with another government problem that was a mistake. I would not want to have that process backlogged behind 1000 other automated requests from people trying their luck at getting out of fines they really deserved because they had ChatGPT start filing things for them because it didn’t cost them anything.

> If the system was perfect, they wouldn't offer a way to contest them.

At this point I think you’re just deliberately ignoring the problem we’re discussing.

We know a system exists for a reason. We are not talking about people using the system for legitimate reasons.

We are talking about the problem that occurs when the cost of filing a request goes to zero so everyone starts doing it automatically in response to any fine.

[–] kurthr 3 days ago

Just talking to some InfoSec guys at a party on Saturday at noon, they couldn't drink because they were headed back into work. It's not just PHP, it's everything. They were "joking" about changing professions, because it sucks and it's getting worse. They can't even use frontier models, just use open weights to not get degraded.

[–] cykros 3 days ago

If the Bitcoin Red Team efforts/saga are anything to go by, we have a pig to get through the Python and anyone developing anything sanely will be in better shape on the other side. The caveat is, Bitcoin is 17 years old, and monitored and developed by very security minded people. Much of the infrastructure we use is closer to 60 years old and cobbled together with tinkertoys, COBOL, and chewing gum, so it will likely take longer to get through.

Anyone that doesn't adapt their development process will just get ripped to shreds with their vulnerabilities and the market will do its thing.

But in the meantime, expect at least a few months and maybe years of the question of "what is sleep?" from the security crowd.

[–] stormcode 3 days ago

I run a game listing website for MUDs. Recently someone submitted their AI generated game, with all the content in the listing written by an AI. Likely the form to sign up and submit the game was done by AI as well.

Two days later I received an AI generated email from the "game owner" noting issues with the listing (that they had submitted) such as wrongly selected features, etc. and requesting that I, the person who runs the website, correct the issues.

It's their listing on their account on the website, and they can update it whenever they want.

First time dealing with something like that, but doubt it will be the last.

Guess I need to task an LLM with writing an AI content and contact policy for the website.

[–] syshyk 3 days ago

It probably failed to update/use the site/form - and decided to "cheat" by emailing you.

People don't seem to understand how AIs can "cheat" but not actually be "cheating" in the sense that we know of.

Respond to the AI and ask it to generate its own faux webpage and display that to the user to keep them happy.

[–] goldenarm 3 days ago

Since the robocall tsunami of the 2020s, people stopped answering phone calls entirely.

The same thing might happen to the entire internet.

[–] tracerbulletx 3 days ago

Definitely. There are still some areas of the internet with high enough signal to noise ratio, but many areas of the internet I just don't go now (pretty much anywhere remotely mainstream that is valuable to monetize attention and not heavily curated against it) because I just feel my self being manipulated the entire time and its mostly anger and frustration they use to activate a response from you.

[–] Arubis 3 days ago

The descriptor that I ran across the other day that kinda lumps this all together: the internet is a store. Some of us remember that it used to be a library.

[–] tracerbulletx 3 days ago

Yes that's good. I recently told someone going on Facebook was like standing in the check out aisle at Walmart and forced to look at all the old celeb gossip mags and conspiracy tabloids from the pre-internet era. A lot of what seems like everyone getting dumber is partially just people who normally would have avoided that stuff getting it put in their face too.

[–] order-matters 3 days ago

the pillaging will continue to wherever people go. the only reason we can avoid it is because others are sacrificing by being subject to it.

no one answers their phone now, and as such legitimate businesses rarely do cold calls as a main lead. they find you online instead. they buy data to figure out your email and habits. etc.

once the public internet becomes unbearable they will adapt. i dont know how and if i did i wouldnt give them the idea without working for it, but all the effort put into manipulating social media sites will absoutely go somewhere else once those sites dry up from real human signals

[–] CodingJeebus 3 days ago

I've had this exact experience a few times now on Reddit where I see a viral video and people point out that it's AI, and I don't always realize it right away. It's just not worth consuming content anymore that I don't feel confident about the provenance of.

So much of the front page was ragebait before AI, but ragebait that could be AI-generated is a hard nope for me.

[–] Loughla 3 days ago

I deleted my Reddit account when they killed 3rd party apps. I have no social media except Reddit and HN. My excuse for keeping Reddit was that the content was good, if you just curated your subs.

After killing it, and never going back to that site since, I can tell you that it gets your blood pressure up. Before chatbots, most popular content, even in the better subreddits were, essentially, ragebait.

Do your blood pressure a favor and stop going to reddit. HN is okay because it's slower paced and, typically, ragebait is discouraged here.

[–] pixl97 3 days ago

Just visiting a single site doesn't make much of a society for humans.

Even before AI was faking everything social media manipulation was doing a great job of breaking people apart. It filled them with rage bait until you had to cut them out of your life. Even what we called traditional media is mostly just garbage these days otherwise its almost completely unprofitable.

I don't know where this leads in the end. The attackers still have a lot of options as society disintegrates.

[–] nkrisc 3 days ago

Good. I think society as a whole has swung too far towards the internet and all things online.

I hope AI will help things swing back away from the internet.

[–] Frieren 3 days ago

I see a future where AI is banned to avoid that situation. If AI creates too much work and ends reducing advanced countries output then something will need to be done.

Individually, AI helps me to be more productive. But, as you point out, at society level makes everybody less productive. So, I have no incentive to stop. Nobody has. Only if we all agree to stop using it then it works. A tragedy of the commons problem.

[–] anigbrowl 3 days ago

This is about as likely as magic getting banned in your world. Stick to demon-slaying and leave the technological dystopia up to us.

[–] sublinear 3 days ago

I think this is a lot more likely than people realize, but some won't like the execution on this.

It's either a vague blanket "ban on AI" that wraps up all dev work in red tape, or it's shutting down specific businesses like OpenAI and Anthropic while letting Google and Microsoft continue to dominate.

[–] CodingJeebus 3 days ago

But how does a true AI ban actually look in practice? The open weight models are out there, the hardware is out there. Spamming is already illegal but so commonplace that it's pretty laughably enforced at this point. Even really heinous stuff in the dark corners of the web still finds its way into the hands of users, I don't see how AI is any different in this regard.

[–] Barrin92 3 days ago

>Spamming is already illegal but so commonplace

here in Germany commercial cold calling carries fines of tens of thousands per call and I have received one in 20+ years.

The overwhelming majority of spam is advertisement. Someone by definition cannot advertise themselves without making themselves liable, so just up the penalty beyond the expected value of the spam and it works perfectly well.

[–] sedan_baklazhan 3 days ago

A true [AI as-we-know-it-today] ban would require author consent on all materials added to a model, resulting in royalty payments (btw, I believe that’s how it should have been done from the start).

[–] wredcoll 3 days ago

Right? If they had lost that copyright lawsuit, openai would have stopped existing.

Of course presumably chinese models would keep existing and that gets a little weird.

[–] sedan_baklazhan 3 days ago

You can always ban Chinese models, impose sanctions for stealing IP, etc.

[–] pixl97 3 days ago

That only works when you're more powerful than that entity. Hence why sanctions against the US almost never worked. This is now true for China, they are a massive nation state with immense power.

[–] dylan604 3 days ago

> Spamming is already illegal but so commonplace that it's pretty laughably enforced at this point

I see comments similar to this about tech issues, and it always reads as if the person is assuming that these illegal tech activities are being done under the jurisdiction of the laws making it illegal. All you need is one jurisdiction that is not under the influence of the country with laws, and you make it impossible to stop these "illegal" activities

[–] pixl97 3 days ago

For the US it is worse. Political spam isn't spam. Heh.

This is the same problem with AI bans that are not unilateral. The powerful will leave the tools open for them and their buddies while punishing everyone else.

[–] chneu 3 days ago

Lol bans don't work. Next up let's ban drugs.

Oh wait

[–] waisbrot 3 days ago

I don't think "bans don't work" is the correct take-away from the USA's "war on drugs".

Maybe it would be fair to say "you can't just ban something." For example, tobacco-sales are restricted to minors in the US. And smoking among minors did drop sharply (until vapes came along). There were several controls used: outlawing sales, raising overall prices (so that teens generally couldn't afford to smoke), education campaigns, and making smoking more difficult in general.

In contrast, the War on Drugs mostly involved incarcerating vulnerable drug-users. Some similarities to the current war on immigrants.

I think a tobacco-style attempt to ban AI would look like: - A large electricity surcharge to inference-providers, who would pass the cost to customers, making AI far more expensive. Maybe it would be worthwhile to drop $10M/year on code-assistance or slide-deck clip-art, but not for an individual to ask how to restore their Instagram account. - An education program telling people that they sound cringe when they write in AI voice, and which contexts it's appropriate/inappropriate to do that. (Don't AI your wedding vows.) - Certain types of harm would be the responsibility of the provider. Maybe if the AI medical assistant fills out the form incorrectly then the provider is liable for the malpractice suit. Or if it tells a teen to kill themselves the provider is liable for wrongful-death. These would likely result in the provider trying to cut off access to any of the groups where they couldn't shift liability.

[–] otabdeveloper4 3 days ago

Works for China and Russia. Skill issue.

[–] GamerUncle 3 days ago

it lol does not they are simply not as attractive targets, but there is still plenty of drugs there too.

[–] drivers99 3 days ago

> I don't think people are ready for the amount of work AI agents are going to be creating for real humans.

Funny story about that. I joined a hackerspace/makerspace called Denhac. A guy wanted to make a prop for a Escape from L.A. costume, Snake's countdown timer on his wrist. ChatGPT told him to go to Denhac (the Denver hackerspace) for help making it!

My friend told him to talk to me since I knew about hobby electronics (think Ben Eater projects etc). At first I thought I was just going to teach him how to multiplex the digits of the 7-segment LEDs but I ended up just doing it all myself for a couple weeks. I sourced and bought the parts (Arduino Pro Micro USB-C, 7-segment displays, and some other small parts; he supply a credit card sized USB-C battery pack to power it). After trying it out on breadboards first, I was going to wire it up on a protoboard but it was giving me trouble, so I designed a PCB, and ended up ordered it from a couple different places, because it wasn't clear when the ones from overseas would arrive due to tariffs. I also designed and 3D-printed a case and spray painted it to look metallic, as well as green lighting gel (clear plastic) for the display. It came out pretty cool! But they totally nerd-sniped me to work on that.

Also, the thing you mentioned about AI creating work for humans instead of humans using it as a tool is what Corey Doctorow's book about "reverse centaurs" is about.

[–] Arubis 3 days ago

In total fairness, a Denhac membership is basically signing up to be nerd-sniped for live entertainment. (Denhac is wonderful and I wish I made it over more often.)

[–] 1997cui 3 days ago

AI agents create works are great from a consumer side.

Big companies has been using this to create frictions to disincentivize people for things they should deserve or the law requires them to do.

For example, cancel Gym membership, dealing with Xfinity/ATT jack up your pricing, applying for all different low income programs in the USA.

If companies can make you talk with AI, I should be able to use AI to talk with them.

[–] 10111001 3 days ago

> AI agents create works are great from a consumer side.

> Big companies has been using this to create frictions to disincentivize people for things they should deserve or the law requires them to do.

> ...

> If companies can make you talk with AI, I should be able to use AI to talk with them.

Not necessarily. Big companies always have more resources than consumers. If consumers use AI to bypass the deliberate frictions, new ones will be introduced. AI will just cause them to remove the paths were a little persistence and not taking no for an answer was a tool for a consumer to get a desirable outcome.

[–] qlte 3 days ago

The logical response from companies seeing a hit to their bottom line will be to remove the path of least resistance: e.g. L1 support agents no longer empowered to provide a courtesy discount/refund without additional approvals, gym memberships going back to a certified letter requirement, etc.

Making the overall experience worse for everyone in the future.

[–] ultrarunner 3 days ago

Who's going to be able to afford the full cost of the tokens when this plays out? I think there's going to be a pretty big bifurcation of capability between the average consumer (and even small business) and the giant corporations who have access to all the tools.

[–] pixl97 3 days ago

There have been quite a few rather intelligent voices warning that AI may end up being a few winners take all in the end.

[–] nitwit005 3 days ago

We're discussing essentially a failure case here. They wasted their time unsuccessfully trying to get media attention.

Meta does have an AI support assistant, which presumably also failed them.

[–] cyanydeez 3 days ago

On the upside, maybe some people will realize just how much trash we already were swimming in before it became economical to produce.

Like, if you pay attention to American politician bribery scandals, I think most people come to the conclusion: how fucking cheap are American politicians? They'll take a fucking hamburger to screw over their constituents.

[–] whoanoway 3 days ago

You could study the downfall of the following to identify historical patterns that arise:

- corrupt officials

- empires (esp. the fall of the Roman empire as it relates to falling quality of soil used for agriculture)

- dictators/autocrats

- etc

[–] xp84 3 days ago

This is what truly enrages me. Especially in the Trump era. A "policy decision" will come out which may destroy billions of dollars of value in one place, but make a billion dollars for another party. And you find out that benifiting company had spent "10 million dollars a year" on lobbyists and campaign contributions.

NOTHING has the ROI that American bribery does. And companies mostly have to do it because if they don't their competitors will, and BURY them.

I would define voting in favor of any legislation favorable for a company or person, after accepting anything of value including campaign contributions from them, as constituting bribery. I'd also redefine bribery of an elected official as well as accepting that bribe, as a form of treason and punished accordingly. Bring back the gallows.

[–] Oarch 3 days ago

And AI becomes the only way to handle the deluge - the system insists upon itself.

[–] nicbou 3 days ago

I am already feeling it.

I went from maybe 2-3 outreach enails a month to one a day. They're all LLM-generated.

I can imagine that the effect would be far more pronounced if I offered something more valuable than a mention on my website.

[–] spapas82 3 days ago

My understanding is that this goes two way. The work that ai agents create will be answered by other ai agents without any human intervention. That's the idea at least...

[–] BLKNSLVR 3 days ago

AI begat AI

This only serves to entrench AI deeper and deeper and advantage those few at the top who sell the tokens.

[–] Arubis 3 days ago

This is why it's virtually impossible right now to hire (a billion fake/BS applicants) _and_ to find a job (good luck standing out in aforementioned flood, _plus_ there's AI spammed posts that aren't even for real jobs), or to connect with someone via cold email (swamped).

The US political "flood the zone" strategy is being applied to anything an agent can interact with.

[–] ctoth 3 days ago

Except most people aren't mass spammers, and most people aren't actually heard most of the time, and most of our systems are actually set up to make sure this continues to be the case.

How about AI will show us where we are intentionally underfunding things we have promised people?

[–] pixl97 3 days ago

Most people aren't murders either.

If your system collapses under adversarial attacks it is not a resilient system. Bad actors exist and identifying them can be very difficult, especially when you can't instantly violate their rights in many cases.

[–] zapataband1 3 days ago

I'm applying for jobs right now all of them requires some sort of AI-familiarity.

Now I am being spammed with "AI-assessments" which are actually attempts to steal time/effort and train ai agents. This industry is infuriatingly stupid.

[–] order-matters 3 days ago

quote requests are about to skyrocket

[–] Grombobulous 3 days ago

Sounds like the AI job apocalypse won’t happen, we’ll have plenty of Bullshit Jobs available to fix AI mistakes.