Hizonner

> These laws exist because social media is extremely damaging to children,

These laws exist because people will believe any nonsense if they see it repeated enough times in news headlines.

"Social media", as presently incarnated, depending on what definition you choose, may be somewhat damaging to some people, including some children. That does not justify ghettoizing all children.

It also doesn't justify creating (yet another) pervasive tracking infrastructure for both children and adults, but that's the part everybody alway harps on. Cutting off teenagers from society as a whole? Nobody's willing to say that's bad.

Well, it's bad. It's especially bad for the ones who are getting the least support at home.

> destroying their attention,

Maybe. Strongest among the claims. Applies to adults, too; there's no sign that adults are one tiny bit more resistant. Best addressed by structural changes to the platforms.

> exposing them to online bullying and extortion,

Occasionally happens. Occasionally happens offline, too. And in online venues that aren't "social media" unless you have an insanely broad definition. One does not lock people in a vault because of that.

> and showing them horrific and traumatic content.

Oh, get a damned grip. You see a picture. Which you probably sought out. It won't kill you. If you're so sensitive that it really gets to you for the long term, your problem is your capacity for "trauma". Something else will "traumatize" you up to whatever that capacity is.

> Can anyone suggest a better way of protecting kids, other than age verification?

Restructure the platforms. For everyone.

> In a trade off between child protection and online freedom child protection will win.

Well, yes, when the "tradeoff" is being made by hysterical idiots. Which admittedly is what tends to happen.

> If I crash without a seatbelt on and die, my going through the windshield harms only myself.

Mostly, yes. Whereas if you fail to get vaccinated and therefore spread a disease, you are harming others.

Well, the assholes currently screwing up the the USA got there by decades of miscellaneous political maneuvers, both fair and foul (most of the fair ones done by an earlier generation), culminating in actually getting elected (basically on a platform of then acting unconstitutionally, because that appealed to enough morons). And the people they replaced also got in through politics.

What they did not do was to sue their way into power. I mean, yes, they used the courts at a few key points, but that wasn't the core of it, and the smart money says they could have done it without, say, Bush v. Gore.

The new court approaches of the 1950s through 1970s were a product of politics way, way more than a driver of it, and so is the present reactionary judicial backlash. In fact, the biggest thing I'd say you could argue was the courts leading, Roe v. Wade, worked for a few decades, but at the same time set up a ton of resentment that was later exploited to help blow up the whole system around it.

And if you go back far enough, you run up against a violent revolution, also not conducted in court. Although even there it's important to remember that revolutions invariably fail if they don't have huge political support first.

So, if you want to actually do something, go elect some politicians who will clean up the mess. By the way, that doesn't just mean going back to the way things were one day before Trump. It means fixing the long-term institutional decay that let Trump and his manipulators cause so much chaos when they happened to win an election with honestly not overwhelming support.

[By the way, I need to edit this: This particular authoritarian move is relatively bipartisan and represents an attitude that's become depressingly common all over the planet. Nonetheless, if you want to do something even about this, the answer is still political.]

Look, a lot of people make the mistake you're making.

Not every unjust, stupid, or evil thing is illegal.

Even when something is illegal, that doesn't mean you have standing to challenge it in court, or that a given court has jurisdiction to do anything about it.

Courts (theoretically) follow rules. They can't just randomly set things aside without some basis in those rules. Lawsuits are not a magic universal remedy.

You could definitely argue that courts don't always follow rules, and that the Trump administration is doing everything it can to make that worse, but the changes they're making aren't going to work in your favor, because those changes are in the nature of "we can do whatever we want, and fuck the courts if they don't like it".

Sometimes. Your point?

I grew up in the phone book age. We had one phone with a really long cable, but it wasn't long enough to take it with me everywhere I went. And, as you point out, nobody had robots to call it, either.

The government is allowed to create regulations that harm people economically. Not much money to be had by instantly losing that fight.

> Email doesn't support the features that those messaging platforms need to have, such as recalling messages.

"Need".

> the right way to build defensive/offsec/malicious tooling with AI isn't by using frontier models ... but run a serious of agents on tightly scoped tasks

The right way to do it is to run a series of agents... many of which are nonetheless built on frontier models (and nearly none of which are built on some local 27B Qwen variant...). One thing the latest models are good at is orchestrating other agents.

I still don't think "enterprise" customers' data have enough training value for an "over-eager PM" to bother with them. An over-eager PM obsessed with AGI, no less. I'm pretty sure more training on corporate slop is not the path to AGI (even less so if you want "aligned" AGI...).

Also, while I do agree that Anthropic's internal controls are unlikely to be on the level of AWS's or Azure's. I'm pretty confident they're good enough that random PMs aren't going to get access to things like that, especially for use in formal projects. Especially since "safety" is Anthropic's other obsession, which means "safety" data are going to be watched.

But anyway, we seem to be agreed that retaining stuff that used to get flushed early is a risk, and every copy is a risk, and sending it to more companies is a risk, regardless of the fine points of how things might go wrong.

You commented on this very issue when it first came up 4 months ago. If I remembered that, so should you. I mean, I'm prepared to believe that you did not think on that occasion, if you want to confirm that's what you mean...

If you don't care about AMD, why are you white-knighting AMD and defending AMD's bad behavior?

But, hey, OK, let's not make it about AMD specifically. It doesn't matter what any company thinks the purpose of its program is, nor does it matter what scope any company unilaterally decides to set for its program. What the outside world is going to see is whether or not you ignore security bugs. Your weird arcane internal policies, justifcations, and "scopes" are irrelevant. And, although I don't honestly care much about "security researchers", you can't really expect them to keep track of your private set of scope rules either... assuming you even tried to tell them the rules in advance to begin with.

... which is why the rest of us should give them, and those who operate them, zero respect.

Nobody but AMD gives a fuck about AMD's internal policies or motivations.

Why would I, as an enterprise customer, care about what queries they answered for anybody else?

Well, to be fair, what Anthropic is actually doing is downgrading anything that could possibly be related to security in any way at all, good or bad.

What they're then trying to do is to use "user is associated with some big Establishment organization" as a proxy for good intentions, and removing the filter when they can establish such an association.

Which is of course blind reliance on a completely untrustworthy signal, prompted by truly idiotic levels of trust in Authority(TM). But it's a different kind of wrong. I do think they understand they can't tell from the query itself.

> I used to be able to tell my enterprise customers something simple, that I really believe: "We use Anthropic models via Bedrock/Azure, therefore we are guaranteed that your data will not be used for training models."

They claim they're not using it for training, only for "safety", and in fact I believe them. If you think they're lying, then why didn't you think they were lying about zero retention before? And "don't throw this in the training bin" is a relatively easy policy for them to get right. Especially because, no matter what your "enterprise leaders" tell themselves, your queries probably have close to zero real training value.

What I don't believe is that they can guarantee it won't leak to non-training parts of Anthropic, leak to or be stolen by outside actors, or be coerced out of them. That risk comes from creating the record in the first place, and that is the problem.

Yeah, like the weird idea that those programs are intended to in some way reduce the number of exploitable bugs actually out there.

The annoying thing about this kind of thing is that it's still possible for the equipment to last a week on the streets.

There has to be a way to solve the coordination problem. One person going around destroying stuff is a "notorious vandal" who can be hunted down and made an example of. Ten percent of the population doing it is game over for any scumbag who wants to deploy this crap.

... but you were silent when they did it to consumers...

"Consensual"?

The only reason I gave up resisting and started writing any significant code in Python at all was that it got some kind of type system, and thus became less unpleasant to code with.

"Pleasant to code with" does not describe getting "AttributeError: 'NoneType' object has no attribute 'foo'" 25 levels deep in a stack trace already obfuscated by dynamic object-oriented nonsense. In production, because it's an unusual case and testing missed it. Not that test cases aren't way more work than types anyway.

Yes, but unfortunately Python has invaded everything, and one must adapt.

Python is going to be preinstalled on almost any machine I use, with a reasonable assortment of libraries. And even if they're not preinstalled, the libraries I want are likely to exist. They'll have unstable APIs and weird quirks, and I'll have to take my choice of bad packaging systems to install them, and everything will just generally be a pain, but they'll exist and largely work. That's not true for any language I actually want to code in. I mean, I'm not going to deny that Python is better than shell scripts or (usually) C.

It's not like it's a pleasant language to code in, especially if you actually want to use the type support, which is weird and irregular and keeps changing and has to work around fundamental design problems at the core of the language.

The conference invited Jay Bhattacharya (the NIH director), who I guarantee was not going to give a technical speech, since beyond an 30-year-old, never-professionally-used generic medical degree, he knows approximately dick about diabetes or anything relevant. These guys were handing out copies of an article outside that room. He didn't show and sent a substitute, who was surely going to sing the same song.

By letting NIH political operatives in, the ADA made the conference itself a political event.

> The article linked doesn't even say what exactly they were protesting (beyond a rather vague "attacks on scientific research" which could mean a lot of things).

It does mean a lot of things, and the Trump administration is doing roughly all of them. Sorry about your lack of basic knowledge of current events.

I would think it would be hard to hit the lens.

The problem with that view is that when the "strong legal basis" is not codified, and codified in a way that nonspecialists can at least vaguely identify and understand, it gets a lot easier to get away with ignoring it. Which the UK has been going hog wild doing in the last 20 years or so.

I am not saying the US is better in practice. The bottom line is that authority worshippers will take whatever liberties they can get away with in any system.

There are no "ROM private keys" in Qualcomm or most other chips. The root of trust is fused in by the OEM. Apparently the exception is Apple.

They would have to individually steal keys from every OEM, in GraphenOS' case meaning Google. Then they'd have to do the right dance to fake the right stuff to satisfy the Secure Element(TM) and get it to let them use the data encryption keys. Which, by the way, I believe requires forking over a hash that may vary among individual phones; you have to know which version of the appropriate stage you want to fake.

... and you'll excuse me if I'm skeptical of your confident statements about what TLAs do or don't have access to, especially when you start talking about keys that don't exist.

> It's not your role to decide or interfere in the politics of other countries where you are not a permanent resident. Think of it like you being a guest.

Well, that's a pretty damned repulsive view.

Countries, governments, whatever, don't have a right to just do whatever they want to their citizens without anybody else noticing. All individuals morally outrank all institutions.

Time to fire up the chaffing. Or the pitchforks and torches. Either one.

Advertising needs to be over now.

Well, yeah, but spammers still aren't going to care. We are not talking about decent or reflective people here.

Don't be an extremist. Let's compromise. We'll only eat one of your children. We'll even let you pick!

You mean the EU's fake snake oil zero-knowledge approach.

They've standardized two systems under the banner of "zero knowledge". One is actually zero knowledge. The other is trust-me snake oil. Guess which one is getting deployment.