There is no need for it. Sorry. Certainly not government mandated, which is the worst part.

YOU have no need of it. You’re projecting your experiences onto everyone else. Maybe you’re content only hanging out in limited forums where you have to know someone to join. And those forums can continue to operate without any human verification. If you don’t care whether the spaces you occupy are infested with bots, or can manage with white lists, fine. But most people don’t want to talk to bots. Not everyone has the social connections you do.

I don’t support mandating ID for anything. But you are not the only person in the world. Just because you have no need of it, doesn’t mean no one does. Not everyone uses the internet the same way you do.

“Only join forums where you personally know the people there” is not a scalable solution to this problem.

Respect a diversity of viewpoints and experiences. Your way is not the only way.

That will always be the case for online age verification. There’s nothing preventing an adult from signing up for an account and then just giving the account to a kid. This is true no matter the verification method chosen. Unless you’re going to demand people scan their face every time they log in, and even then there are ways around it.

My goal is not to create a perfect solution. My goal is to create a good-enough solution that is as reliable as face or ID scanning, but without the privacy violations.

There is a clear need for it. There’s value in having online discussion spaces that aren’t just swamped with bots. We really need a reliable way of telling human from bot. I want to find a way to do that preserves privacy to the greatest degree possible.

Do you have any constructive criticism to offer? Maybe a suggested improvement, or an alternative way to accomplish the same goal? What’s your solution to separate human from bot?

Nothing. Who says a solution has to be perfect? If “good enough” is good enough for handling the sale of alcohol, it can certainly be good enough for accessing web sites. No solution to any problem is perfect. You always have to find a balance between preventing the thing you want to stop and the negative effects of your efforts to prevent that thing.

Person A gives an AI a list of bullet points and instructs it to expand the bullet points into an email.

Person B receives the email and asks AI to summarize it into bullet points.

An ocean of water could have been saved if the Person A had just emailed the bullet points directly, and the information transfer would have been far superior.

Exactly. I use AI primarily as a natural-language search engine, when searching for things that can quickly be independently verified. One of the best uses for it is when you know a thing probably exists, but you don’t know the name for it. You can describe to the LLM the object or problem in detail, and it will give you a name. You can then take that word and do a non-AI search to instantly confirm if it got it right.

I’ll never use an AI for something I can’t at least verify without an AI.

So you teach something different from the mainstream - cool. But don’t be surprised when students or AI cite others or take an alternative position.

Not every class is debate. Plenty of courses are simply taught by an expert in their field, and students are graded on how well they learn from that professor’s instruction. Is it a perfect system? No. But that’s why we have experts. Some classes exist to simply establish a baseline of knowledge among the students. Such classes cannot be continuously derailed by engaging in rigorous debate about topics that experts in the field consider unambiguous and long-settled. Students are free to form their own opinions and whatever conclusions they want, but they need to demonstrate that they have at least learned the material as taught. You can believe the Universe is 6000 years old if you want. You cannot expect your astronomy professor to spend half the class debating with you about it.

A creationist student doesn’t have to believe in evolution to get an A in an evolutionary biology class. But they do have to demonstrate mastery or at least competence in the principles of evolutionary biology as taught by their professor. If they want to view the whole course as “a silly exercise in memorizing the lies of evolutionists,” well they’re free to do so. It’s a free country. Believe whatever you want. But that doesn’t mean they can just write, “God did it” to every answer on the final and still receive a passing grade.

You’re free to believe or reject anything a teacher or professor teaches you. That doesn’t mean you get out of learning the lesson entirely.

I love your approach. This way, you don’t even have to bust the students for using AI!

It’s often very difficult to tell the difference between a student that uses AI and a student that performs a traditional search of the topic and just copies facts blindly and uncritically. AI writing detectors are simply not reliable enough to serve as the basis for a charge of academic misconduct. What do you do if you accuse someone of using AI, and it turns out they just happen to write in a style that’s reminiscent of AI? What do you do if someone claims they just happen to write how an AI sounds? What if someone says they just used Grammerly, so now it’s ambiguous what portion of their writing is their own? Academic misconduct and AI gets very thorny very quickly.

But your approach? Brilliant. Just give them a paper whose topic you taught them correctly, but you know AI will screw up. No need to affirmatively prove the cheater used AI. Just give the cheater a failing grade on the paper, because it clearly doesn’t reflect what you taught them. They have no ability to later argue that they’re innocent or lie that they did not in fact use AI. Their paper simply wasn’t good enough to get a passing grade.

Honestly, it’s pure genius. As a fellow educator, I commend you!

Yeah, that’s why in an ideal case, I could see a two-tier model developing. Similar to how we handle weapons today. At least in the US, handguns and rifles are pretty obtainable. But anything more than that is heavily restricted. Even 2nd Amendment folks don’t advocate for the private ownership of heavy artillery, FPV suicide drones, armed tanks, etc. We allow anyone to own the things that, while still dangerous, don’t give you the power to kill absurd numbers of people. Even with an AR-15, you’re only taking out so many people before you’re taken down. You’re not a danger to civilization by owning an AR-15.

Computers could work the same way. As AI technology improves, we’ll get a better understanding of the minimum level of compute needed to be dangerous. You then set a limit for unlicensed computing below that level. Unlicensed? Have all the privacy you want. This lets you play games and create text, video, and most any form of expression you want. If your creative medium is as an artist of fully immersive 4K virtual reality worlds, you might have some trouble. But pretty much anything you currently use a computer for would fall under the unlicensed compute threshold.

Above this threshold? Everything is locked down. Everything is logged. Everything is scanned. You’re using a tool that is as dangerous as a pile of plutonium, and we treat it with an appropriate level of respect. Dealing with this privacy violation is just a price you have to pay if you want to use these tools.

Honestly, it’s something where I wonder if we’re doomed to end up there anyway.

I think that the AI companies vastly overstate the usefulness and danger represented by AI models. But the technology is only getting better, and the resources required to run models will decrease over time.

A common scenario brought up are AIs teaching people how to make biological weapons. Can existing LLMs actually do that? I really don’t know. I for one don’t feel like asking copilot, “please give me step-by-step instructions for obtaining and weaponizing anthrax, in steps simple enough someone with only a general science and engineering background can do.” I’m curious what copilot would say. But I won’t be asking copilot that, not because I’m incurious, but because I don’t feel like getting my house raided by the FBI.

But imagine a world where that actually is possible. Forget the claims of Sam Altman. Let’s look 50 years ahead in the future. Imagine being able to ask that question to an open-weight/model LLM, something you can run unmonitored on a home desktop, and actually getting a correct response. Imagine being able to ask endless follow-up questions. A good enough LLM could conceivably walk someone with only a high school education through all the steps needed to create weaponized anthrax from scratch.

Anthrax is something we actually can’t control the spread of. It occurs naturally in the soil and water in some specific locations and conditions. A sufficiently skilled biochemist could go out, collect it, and culture it. And having an LLM capable of giving you step-by-step instructions, dumbed down to whatever your skill level is? Able to coach you through it, perhaps even watch you do it and offer real-time feedback? If you just have to act as the hands of the LLM? Suddenly anyone can weaponize anthrax.

I don’t think existing LLMs are there yet. But what do we actually do if we reach a point where open-source LLMs are? What are we supposed to do when the knowledge to produce weapons of mass destruction becomes cheap and accessible?

A general trend in technology over time is that it has magnified the potential destructive power of a single individual. How can a society survive when anybody has the ability to kill everybody? I just don’t see how you keep society going except by locking down the tech. Maybe require a license for any computer over a certain power. And such computers are heavily surveilled. So you can have a free and private computer, but only if that computer does not make you a danger to everyone around you.

I just don’t know how else a civilization can survive the democratization of weapons of mass destruction.

I have a solution to the age verification problem. There is a way to affirmatively prove someone is a real human adult without invasions of privacy. We can use the same ID verification system we’ve been using for centuries: public notaries.

Governments could hand anonymous cryptographic tokens to notaries. These contain no information on the individual. They’re simply a unique cryptographic token. You can go to a notary, pay a nominal fee, show your ID, and grab one of the tokens (possibly just a code printed on a card) from a large bin of them. You can then use this token to register for any number of sites. The notary themselves does not need to note which cryptographic token you grab. The notary doesn’t even know what token you received. The goal is merely to prove you’re an adult human, not to create a cryptographic key tied to your specific identity.

I would then let people do this as many times as they want. You can get a hundred such IDs. They wouldn’t cost much, a few dollar or Euros. This would be no barrier to individuals accessing the net, but it would make them unsuitable for mass spamming.

They’re correct that this creates cognitive dissonance. The racist, upon meeting a minority member that doesn’t uphold the stereotype, has two contradictory facts in their head:

1. All members of a minority group are bad.

2. This member of that minority group isn’t bad.

This is cognitive dissonance. Both of these cannot be true at the same time. One of these premises must be wrong.

Hollywood imagines the bigot will treat this like a physicist disproving a theory by finding a counterexample. But bigotry is anyhing but scientific. You can’t reason your way out of a place reason did not lead you to. Instead, they adjust their premises to:

1. Most members of a minority group are bad.

2. This member of that minority group is one of the few good ones.

Read more carefully.

cmbabul basically said, “trans people shouldn’t defend themselves, as that would just embolden bigots.” Primecf pointed out that bigots will just make shit up, as Republicans do when they pretend every mass shooter is trans.

cmbabul is demonstrating behavior found among some allies, where they nominally support trans rights, but get really uncomfortable with the idea of a minority group defending itself with lethal violence. Minorities are meant to be protected in a parental way by the majority. Minorities are not meant to have agency of their own, certainly not in the realm of self-defense.

If money is a form of speech, so are bullets.

Maybe we shouldn’t bait-and-switch people by calling something private that is anything but.

This is just victim blaming. These people are victims of corporate deceit, and you’re blaming them for it. What was she wearing? Maybe that contributing to it as well!

That’s the fundamental problem I have with this. I don’t believe in trapping people through gotcha bear traps. If you’re going to be scanning every message for violations of rules, you shouldn’t be advertising that service as private. In fact, they shouldn’t even have DMs in that case.

War is Peace.

Freedom is Slavery.

Your Private Chats have no expectation of privacy.

Read more carefully. I’m obviously not talking about how these online spaces are legally defined. I’m talking about how the companies instruct their users to use them. I don’t blame people for acting in a way that they’re told they can act.

“The private chat has no expectation of privacy!”

Orwellian bullshit.

You are advocating for the death of children.

The policies you support will, on net, kill far more children than they save. These inevitably have an extremely high false-positive rates. For every one actual school shooter you catch, you end up ruining the lives of a thousand innocent children. Many of those will have their lives completely disrupted, many dying early deaths of from suicide and drug overdose.

You are are advocating for the death of children.

Bullshit. This wasn’t a public space.

Imagine I rent you a hotel room. You expect it to be private. Except hidden in the fine print of the check-in paperwork it states that there are microphones in the hotel room and anything you say there can be reported to the authorities.

That is the comparable situation. Quit equating this to public speech.

These companies encourage us to treat these tools as private. They literally fucking them PRIVATE CHATS. Because that will encourage us to use their platforms more.

If this is going to be the standard, we should make it illegal for companies like this to offer private chats as a feature. If the speech is going to be policed this much, we should make it so everything has to be published, in public, for anyone to see, under your real name.

The problem is the tool is sold as as private, but then policed like you’re standing outside a school with a bullhorn threatening to shoot the place up.

I guarantee you if this teacher killed the student people would be carrying pitchforks over nothing being done about the messages.

And they would be fools.