Hacker News Reader

Comment by dom96 | Hacker News Reader

I expect that nowadays many online are speaking with GenAI without even realising it.

It’s already been bad enough that you may be unknowingly conversing with the same person pretending to be someone else via multiple accounts. But GenAI is crossing the line in making it really cheap for narratives to be influenced by just building bots. This is a problem for all social networks and I think the only way forward is to enforce validation of humanity.

I’m currently building a social network that only allows upvote/downvote and comments from real humans.

View on HN

Dead Internet Theory https://en.wikipedia.org/wiki/Dead_Internet_theory

>I’m currently building a social network that only allows upvote/downvote and comments from real humans.

And how exactly do you do that? At the end of the day there is no such thing as a comment/vote from a real human, it's all mediated by digital devices. At the point the signal is digitized a bot can take over the process.

By validating each account with a passport and ensuring that only one account per passport is allowed.

What about people who buy stolen passports on the dark web? Or passport details that get leaked in data breaches

We need a German to chime in with whatever word describes this scenario, when someone suggests an action is not worth doing because of corner cases or inability to perfectly execute a process.

In this hypothetical, let's say we'd tackle the dark web passport market issue when we get there.

hehe yeah, it's funny how many are focusing on the small edge case that makes a certain solution not 100% perfect.

There is also another issue: people can have more than one passport if they're dual citizens. But you know what... I think that's fine.

Great points

In this case, the english word overkill does a good job already.

Passports have digitally signed certificates in them, readable by any device with an NFC radio. It's easy enough to extract that data with a mobile app, and now you have an unforgeable file that can be hashed. Of course whether users will bother to sign up for something like a social network if they have to go rummage around and find a passport, install a mobile app etc, I don't know. But it's technically possible and I've proposed such a scheme years ago. For bonus points do the hashing inside a secure enclave so the ePassport data is never upload to a remote server in the clear.

who is going to upload their passport to use social media?

People already upload their passport for lots of things, why not social media?

> People already upload their passport for lots of things

I sure don't.

> why not social media?

privacy?

For me, sacrificing some privacy is worth it to be in a community with users who I know are real people. If that's not something that is important to you then that's fine.

In my case, right now it's very easy for you to figure out my real name by just googling my nickname. Registering on a website like the one I am implementing won't sacrifice much more of my privacy.

Why would I ever give my passport to your website or anyone elses?

Because it enables you to be a part of a social media that is guaranteed to be majority human-centred. Why wouldn't you give your passport to a website? Don't you already do so for crypto or other banking?

Aside from the practical and technical problems you're greatly limiting your audience. Most people who don't travel internationally don't have a passport. In Europe this might be a smaller number but in the USA and Canada I would guess this is a majority of people. Non-citizens won't have a passport. Most young adults will not have one. Many older people will have let theirs expire.

Yeah, that's the challenge. My bet is that there are enough people out there with passports to make an interesting social network.

Of course, getting someone to share their passport will be another filter. But I hope that I can convince people that the benefits are worth it (and that I will be able to keep their data safe, by only storing hashes of everything).

Ok, so you get some critical amount of 'humans' to share "a" passport. Now you've built an expensive and high quality lead finder for scammers/spammers. You've increased the value floor for people submitting fake passports. Also, how are you paying for verification of the passports?, VC money to get from the loss phase to the making enough to support itself on ads? How are you dealing with government data requests, especially in the case where said data can be attributed to a real human?.

Maybe I'm wrong, but just a social network of 'real people' doesn't seem like enough in itself. What is going to bring people there with the restrictions and potential risks you're creating.

You could very well be right. I'm willing to give it a shot and see.

All I can say is that I personally see huge value in a social network for real people. Personally I am sick of arguing with what are likely legions of duplicate accounts/bots/russian trolls online. I want some reassurance that I'm not wasting my time and am actually reaching real humans in my interactions online.

Success to me is 1000 MAU. There are companies out there that do passport verification for a reasonable fee with a fair amount of free verifications to start with (which will handle 1000 MAU just fine). If the number of users wishing to take part is significantly higher then I will explore either ads or charging a small fee during registration.

I'm still very far from needing to cross that bridge though. Same for some of the other questions you've raised. I'd have to do a lot more research to come to a solid stance of what to do when government data requests come in. But I would guess that there isn't much choice but to abide by the request. If you want true anonymity from the government then this place will not be for you (but then I'd say not many social networks are for you in that case)

I see a lot of value in a network where I can be confident I am talking to real people.

As a user I can't do anything related to the passport stuff and I know many people who likewise wouldn't be interested in doing that, because we live in the states. A more "normal" approach here would be to use one of the government ID verification systems for your state ID. Most of us are willing to expose that information, since it's what you show when you go to the store to prove your age/identity.

I’ll definitely look into this. But starting with them would limit me to the US only (I’m sure other countries have similar things, integrating each would take a lot of work though). I want to start with passports because they are a true multi national document.

I would stay the hell away from any site that wants a copy of my passport.

As for a new turing test there is a magic word that bots are not allowed to use that guarantees you're talking to a human.

There is no such word. Even if there is it would be trivial to create a bot that avoids this.

Passports are the only practical way to ensure that I’m talking to a human. Do you have a better idea?

If I think you're a bot I'd ask you to say the word. The word exists and it works every time. No one is going to train a foundation model just to say the word, it's too expensive.

Prompt injection can be mitigated but not this prompt rejection.

You're so delusional, mate. There is no such word.

That is your loss then.

The idea of uploading my passport in order to talk on a social network is quite delusional.

I would say LLMs have told me more interesting things than any human has in the past year and it is not even close.

I suspect at some point, a new structure will be figured out that it doesn't matter if you are talking to a human or LLM. If that doesn't happen, at some point I will probably just stop trying to talk to humans online and just talk to Claude or whatever the strongest model is of the moment.

No passport, but perhaps face picture that is also encoded with on device verifiable token.

I think there is actually a use case for blockchain (don't pile on!) for this. I have a vague idea of a unique token that goes on every camera and can be verified from the blockchain. If a picture has the token you know it's real... like i said, it's vague idea but i think it's coming

No need for blockchain. All you need is this: https://contentauthenticity.org/.

The problem with this is that it's still easy to forge.

I'll certainly consider playing with ways to identify human uniqueness that don't require passports. But passports are the most obvious route to doing so.

Seems like it's sorta what I'm talking about? Hard to judge because they need to work on their communication skills. It reads like techno-corpo-babble and I'm a professional software engineer.

This seems more targeted at us: https://opensource.contentauthenticity.org/docs/getting-star...

How do I know that, you will handle my passport data with care? Banks I can trust(despite numerous leaks), you as a random social media or online service with zero regulation, I won’t. Plus this opens up immense ways to sue you for collecting unnecessary data and personal information, unless you are massive and have an army or lawyer or have a KYC requirement.

I plan to outline exactly what data I store. I don't plan to store raw passport number/name details, rather a hash so I can verify the same passport isn't used twice.

So even if the DB leaks no one (except the government) will be able to tie your real life identity to the account.

In the US there are a lot of companies that do these kinds of identity verification for you. I had to do one when I was setting up a new Steam Developer account and many use them for regular day-to-day verification like for work or when submitting important government requests.

It would seem a lot better to just partner with an existing company that takes care of that part of identity verification. Your job is still to compose all of these signals and metrics and distill them into a simple "everyone is human" network, but the actual job of being a passport jockey can be avoided IMO.

Yep. I’ve considered doing it myself but it would require setting up an app for iOS and Android. For now relying on existing services will be fine. If this takes off I might set up my own passport verification service.

Why not require a non-voip phone number instead of passport?

SIM cards are as cheap as £1 where I live. £1 per account is nothing and I wouldn't be surprised if there are ways to get unique numbers for even less.

There are lots of ways to get verification codes for any number of phones for absurdly cheap. They run large farms (or hacked devices or shady apps, who knows) and can provide a service where you get hundreds of SMS verifications for a few bucks a month.

I agree with the other commenters that you'll face a lot of challenges, but personally I hope you succeed. Sounds like an idea worth pursuing.

Thank you :)

>I’m currently building a social network that only allows upvote/downvote and comments from real humans.

What's your method for detecting real humans?

Also will your social network allow bots labelled as bots?

See discussion above for details, but in short: by requiring user passports.

Yeah, I’ll probably make it possible to set up bots that are clearly labelled.

Good bot (/s)

I don't know that "real humans" is good enough. You can do plenty of manipulation on social networks by hiring lots of real humans to upvote/downvote/post what you want. It's not even expensive.

Yeah. But the cost is significantly higher than ramping up GenAI to make you thousands of accounts.

There is no fool proof solution to this. But perfect is the enemy of the good. Right now social media is pretty far from being “good”.

Reply on news.ycombinator.com