Dylan16807

I think any analogy with physical theft is too different from data to apply to this comparatively subtle case. Especially when we get into the details of just using the output of the model to train on.

For me, when it specifically comes to copying, I don't think it's bad to copy a copier. (And by that I mean Anthropic has no valid complaints against Moonshot. Any valid complaints from anyone in the original corpus are valid against both of them now.)

In this way, it is different from literal theft. Stealing money/objects from a thief and keeping them is not justified.

I mean yeah.. An argument is typically something pretty different from a "point"... Idk why that seems so crazy?

Because it feels like you aren't even trying to understand me if such a tiny wording difference makes you skip over what I'm actually saying.

I wouldn't even say I was making the point.

I know! You quoted them, and I took issue with a thing in the quote!

I was, despite its seeming futility now, trying to make a point about the article to you. It seems willfully misread to take my last post in any other way.

I can see you saying things about the article, but I'm not disagreeing with you (except to the extent that you're defending the paper), I'm disagreeing with the authors.

I don't know why you are talking like I did this experiment?

I don't know what I said to gave you that impression. I keep pointing at the line you quoted when I explain my complaint.

My first post in this thread didn't include a single word of yours. It was only the quote.

There is literally no delegation in the way (I think) you are meaning, there is the contrived presentation of different tools for different control groups.

Forget what you think I mean. Forget what words I used.

They said "precisely so any reduction in judgment could not be explained as sensible delegation to a reliable tool". They were trying to make a particular explanation impossible here. That's what I'm referring to.

They wanted to avoid that explanation, and I'm saying they failed to avoid it, possibly even backfired and did the opposite of that.

There's no point in nitpicking the words I use to try to refer to that sentence. Pretend I'm just drawing a big arrow pointing at that sentence.

I am not over here trying to argue for anything other than common sense and the ability to understand what the researchers were doing using even a oz of intellectual charity.

I'm also arguing for common sense and understanding what they were doing. And I'm saying that in this specific aspect of what they were doing, it didn't work.

I'm not judging the entire experiment here, just that one aspect.

Is that, too, impossible now? Where is the freaking fire?

Is me insisting I see one flaw acting like there's a "fire"? I don't think so.

At the very least, consumer connections explicitly disallow "Commercial Use".

And that's really bad! Far more than reselling, you should very much be able to run a website from home!

On the other hand if you want to resell 2% of your bandwidth the ISP shouldn't have any say in that.

Sometimes it's botnet, sometimes it's just accessing netflix without hitting big IP range bans.

If there are proxy apps that only do the latter sort of work than I'm actually in favor of them existing and being widespread.

The article, and the paper it references, is not an argument or "proof" of anything. This is not a theorem. The point is, given access to the same relative information as anything, there is (supposedly) a greater trust, there is less I-dont-knows with something in the form of a chatbot vs something else.

...Are you saying there is a difference between an "argument" and a "point"? And you accuse me of missing what people are saying..

Okay, they were making a point about how people delegate. They wanted to remove a confounding factor "so any reduction in judgment could not be explained as sensible delegation to a reliable tool." But because of how people judge things, they failed to remove that factor, and possibly made it worse.

It's not even, really, about "delegation" itself. It's just studying the supposed correlation here between uncertainty and one certain form of a tool.

But they decided they cared about removing the "sensible delegation" explanation. I'm not imposing on that on them. They thought it was important to remove, and they did something that doesn't remove it at all.

It could theoretically help to flag those images. But that kind of submission should be anonymous. It shouldn't ruin the ability for someone to get therapy.

Mandatory reporting makes sense for situations you are connected to. And even then there's presumably good reasons not everyone is a mandatory reporter. This goes way beyond that, mandatory reporting once removed for someone that doesn't know a single person involved.

Hell, reporting someone for that doesn't even guarantee the images get looked into! If they didn't save history they're probably not feeling like going back to the site to demonstrate. Similar if it was sent against their will and they deleted it right away.

That problem just requires there be big GPUs to hack into. The number of those sitting around will keep going up. Very much not scifi.

A couple terabytes aren't that hard to move around. And you can split a model across many many GPUs if you'll tolerate it being slow. And you can run many parallel threads to keep up throughout.

As if the immediate future wasn't billions of these tasks...

There's only so many GPUs and a lot of them are devoted to patching flaws.

Many successfully improving their own capabilities

I haven't seen much of that. But that also applies to the ones on defense.

And more flaws are probably going to take increasing resources to find.

Presuming that the hacking program that is breaking into other computers could likely get a copy of its own files is not "science fiction". Or it could just be given them by the owner!

No, what I'm saying is: If your password was good enough for a quality KDF, then improving it to use a 50x worse KDF is very easy. If the worse one requires you "use an AES key", then the good one pretty much also required you to "use an AES key". If the good one didn't require that, then the 50x worse one also doesn't require that. Generate a single letter and slap it on the end, or human-pick two letters.

50x is not the decider between allowing a good password and allowing a bad password. It's a tiny little nudge.

Edit: To put "little nudge" another way, if you have a 1-10 scale of password quality, most steps in that scale are going to be more than 5.6 bits apart.

Random is important, character set not so much. And yes when I say 2 characters I mean random ones. Even 26x26 is plenty to compensate. Or you can add one random word.

You can compensate for not just 50x but 500x by adding 2 characters to your password. If that's the speedup factor, it's not itself a big worry.

Reporting abuse the client was involved in has a compelling reason. That's different from hearing their client saw a picture of the abuse of a total stranger.

Well if you're just talking about surface level feelings with almost no connection to reality then I'm not sure why you brought them up. Okay people think that when they don't think. You can't draw useful conclusions from that.

Requiring a zero, okay. By the time you also require that none of the other numbers are zero, I don't even call that a trit anymore, it's a bunch of weird constrained values.

Even if someone went browsing for it, yes that's illegal but there's no benefit in their therapist reporting them for just visiting terrible websites.

But also there are definitely ways to get accidentally exposed. That's an absolutely awful thing to call the cops over.

The average person has given up on RAM prices coming down.

It's been less than a year. That's way too fast to give up.

Anyway TSMC isn't making RAM.

Who are you suggesting do so? It would be weird to think mainland China would give consistent and non-manipulative pricing to the rest of the world, and any other option is very silly.

No matter how or why? That seems like a terrible mandate.

I've never used AI except for sometimes getting distracted by google search's builtin wrongness factory. So whatever "revealing" you think you found is completely imaginary. Rethink your assumptions here.

So please make an actual argument. How am I missing the point? It's true that someone trusting the AI in this test is not practicing "sensible delegation to a reliable tool". But what actually matters is whether they are practicing "sensible delegation" full stop. There's a big difference between "the subject inappropriately trusts AI in general" and "the specific test setup deceived the subjects". In the latter case, the attempt to remove the "sensible delegation" factor failed.

Edit: And any argument that uses "any LLM they would normally use is unreliable" as a basis is begging the question. If you can just assert that then you don't need to do anything to disprove sensible delegation. But if you can't assert it, the proof doesn't work right. So either the proof is pointless or it's insufficient.

Making a particularly good core is hard. Making a core to the minimum risc-v spec is what you do in a single college course. It's what the ISA was designed for.

An "intersection of INI and TOML" would presumably only have scalar values. TOML allows really complex lines like these:

  data = [ ["delta", "phi"], [3.14] ]
  temp_tagets = { cpu = 79.5, case = 72.0 }

precisely so any reduction in judgment could not be explained as sensible delegation to a reliable tool

That only works if they're experienced with model(s) of that level of unreliability and this is presented as one.

If they're used to a model that's more capable, and think the test model is similar, that's a huge confounding factor all by itself. It's not quite like giving fake credentials to a guy off the street and presenting them as an expert, but it's largely similar.

Your overall assessment of advantages and disadvantages seems like you're comparing paper with backups to software without backups, though. No competent system can have all the records destroyed remotely. And we know how to backup digital systems even better than we know how to backup paper ones.

And as a third option we can have efficient digital systems that aren't plugged into the Internet. (Presumably the Internet could have a copy that's regularly updated.)

Yeah, while online copies are a risk, you can make offline digital copies of important data for a thousandth the price of paper copies.

Put some desktop-size tape robots in several government building closets, and task someone with switching tapes weekly, and you can achieve more reliability than multiple huge paper archives.

To get one, but not to renew it, so that's still pretty rare.

So when I say "every mention" I'm not talking about how many times you in particular did so, it's about it being brought up a lot and this being a completely unrelated post that really doesn't need the complaint.

I don't understand what's horrifying you. Assume "is this consistent with everything I know" is yes here. Now what?

And "what would it take to change my mind" is a picture of the film for most of these. Is there a problem there? There's very little chance these people are going to trust the AI over their eyes, they just don't want to bother hunting down pictures to use their eyes.