ExoticPearTree

Maybe Anthropic is in some kind of trouble and if the IPO does not go the way they dream of, there is a scapegoat they can blame.

Or Amodei is really far gone in his beliefs.

Right, because this is what tge world is missing right now: a black market for AI models.

For all the good things Anthropic makes, it is a very unhinged company.

> He's never going to be hired as CEO by anybody for the rest of his career.

You must be new to the corporate world or incredibly naive.

> What do you think happens if they refuse to do it?

In war time they can be shot on the spot for disobeying orders.

> Either nothing, or if they get sued or prosecuted for not doing it then they get brought in front of a court, and a judge who is doing their job would say that they were right not to do it.

I think you confuse the military with corporate environments. Worst thing that can happen in the private sector is that you get fired.

In the military you can spend the rest of your life in prison if “you did not feel like obeying orders”. The military is built ob discipline and followimg orders, not Gen Z mood swings.

I don't think you understand what "No warranty" means in a software license. It means Anthropic cannot be held liable for results if the software misbehaves and instead of bombing a terrorist compound it bombs a wedding (I chose this example because the US has a tradition to bomb weddings in Iraq and Afghanistan). But in no form does it mean "yeah, I don't feel like doing what you prompted me".

Hence, Anthropic sabotaged the models on purpose to not respond properly to all prompts.

That means that the software was not tested for every scenario. Not that the user cannot try to use it in a way it was never intended to.

If, on the other hand, Anthropic deliberately blocks the user from doing something, then it is sabotage.

Yes, a very small list of NOs. Everything else goes.

Again: if Anthropic deliberately blocks a user from ordering it to prosecute a target, then it is degrading the software capability on purpose.

There is a dfference between “our product can’t do X” and “our product can do X, but we don’t want you to use it to do X”.

Say we have this prompt for Claude at the Pentagon: “You have control over these 50 drones, fly them to Ethiopia and eliminate all possible terrorists you can using the maximum endurance of the drons. In this folder you have all the necessary intel. Confidence rate needs to be at least 70%. Report back when all drones have landed back at the base”

Why is this not considered human in the loop? And let’s say human in the loop will just press Y; what is the difference?

> They're clearly fine with it, but they wanted human-in-the-loop on decision making.

And the Pentagon does not want that. Plain and simple.

And now Anthropic is learning the hard way what happens when you say no to the Pentagon.

It is truly mind boggling that anyone thinks Anthropic can dictate terms to the military.

The passed a law directing the military to round-up and arrest every japanese in the US and hold them indefinetely, and the military did so.

The military will kill anyone instructed to kill. Especially if it is in the body of a law passed by Congress.

Do you think in the military you can just be like “Yeah, I don’t feel it today, I’m going to take s sick day”?

> The US is a constitutional republic, it does not have a system where anything Congress decides is law.

Really now? Then why does it say in the Constitution that it was created to pass laws?

I thought after 60,000 feet MSL a country's airspace ends.

So, who's the UK actually attacking?

Ah. My reasoning, with the normal distribution, was that zero was at the left most end of the curve and with a pretty much non-existent chance of appearing.

> You definitely would expect a roughly equal number of 0s as any other of those numbers since it's uniformly distributed.

How would random numbers be uniformly distributed?

The probability of generating a zero is incredibly low if you use the normal distribution curve.

So it is not necessarily that it doesn't generate zero, they did not run enough times to increase the probability of actually generating a zero.

For example “I have an MBA” and “I know a mechanic”.

When you pronounce the M from MBA, you hear the e in front, you’re saying em-bi-a. But for mechanic, you just have a stromg m.

This is what I mean by “if it sounds like a vowel”. And hence an and a.

> You think it's okay when actual harm and suffering results from a "battle of the brains" via computers, because one party "outsmarted" the other. Sure, it would "suck", but you think the playing field is pure and level, making it a fair contest and any consequence fair game, and therefore it should not be illegal.

I said the fault lies with the administrators of those systems, not with attackers. I really think I never said it should be legal or illegal. I don’t really care if it is illegal or not, hackers are not dettered by the legality of it. Do you think someone in The Gambia cares that hacking is illegal in Canada?

> You are unable or unwilling to engage with the question if and why using a computer and "smarts" to cause harm is different from using strength

For me being smart and being strong are two wildly different things. It is like asking me why I don’t compare apples to oranges. I can’t.

> In any case, that's a very interesting position. I'm curious what you gain from arguing it. Where does that come from? It's possible you're just trolling, but maybe smarts and brains connected via networks are truly special to you. Why?

I am not trolling. I see this, hacking and securing something against hacking, as an “intellectual fight”.

Let’s say you are a 50 year old security admin that gets owned by a 14 year old with a computer. You were beat because the 14 year old one was smarter than you, not that he had more experience or was physically stronger than you. Just smarter.

Now imagine you’re part of a security team and you still get owned. What does that say about you?

I am at a loss on how to explain that when it comes to computer security the fault, in my book, does not lie with the hacker.

Just like when a flaw is found and exploited I do not blame the one who found it, but whoever made it possible in the first place. And in the case that the flaw was patched and a software update was made available, but it was not installed promptly, then the fault lies with whoever did not update the system.

Regarding arguing: I made a statement expressing my position and got mobbed for it. Now I am defending my position.

We can agree to disagree and keep enjoying what is left of our weekends.

Because “an youghurt container” sounds wrong.

English has this thing where you know you wrote or pronounced something correct if it sounds right.

To keep it simple, you use an when the first letter of the next word sounds like a vowel when pronounced (vowel sound) and appplies to ancronyms too. Otherwise you use a.

> Does that extend to medical systems?

Anything network connected that can be reached by an adversary. And, I did not say it shoukd be legal or illegal, just that the fault lies with who administers/secures those systems, not with whoever breaches them.

> How do you compensate someone who's dead?

In some countries where I lived, there were pricelists based on nationality that insurance would pay out in case of accidental death. If you live in a more homogenous country, you can use other factors to determine what the payout should be.

Semantics. If you have a computer, the hacker has a computer, and you get hacked, the hacker is smarter. For whatever values of better/creative/resourceful you want to attribute to “smarter”.

In a nutshell, yes.

My issue is that software security is not taken seriously most of the time because features are more important than spending a little more time on code quality.

The hacker is not the one writing buggy software.

> Both are about preventing harm in many different forms. Software famously has effects in the physical world, that's the reason why a lot of it exists, and why people get paid that deal with software because it makes their brains feel good.

Yes, it would suck if the hospital got hacked while I underwent a surgery for example, and the ventilator stopped working. But if that happens, whoever is doing it is not stronger, just smarter than the people administering the hospital network.

> From your perspective, what makes "smarter" different from "stronger", or "more resourceful" here? Or do you think that if your door gets bashed in, it's your fault, because your door was too weak? Or your head? What if someone outsmarts your physical security arrangements to wander around in your house? Where's your boundary here?

I gave an analogy with chess. You don't have to be strong in the physical sense to win a chess match, just smarter than your opponent. This is how I see the difference.

> You seem to be advocating to have none at all for the software case, and I'm trying to find out about that.

For software, the playing field is level: you use a computer, your opponent uses a computer. But the difference is the other person's capabilities. You can be smarter and you don't get hacked, or your opponent is smarter and hacks you.

Because it is a battle of the brains, like when you play chess. And I believe the smarter one should win. This is why I have this strongly held belief that if you get hacked, it is on you. The attacker was smarter than you, simple as that. So you have to get smarter and become better. Or you get hacked again and again.

I really don’t understand the urge or need to compare software security with physical security.

I don’t care in that scenario, it is the bank’s fault, not mine. It is on them to make it right.

When it comes to computers and software, yes. They need to do a better job not letting others screw them over.

I have this view when it comes to computers and software, not physical property or people.

I have this view when it comes to computers and software, not physical property or people.