Hahahahahahahahahahahahahah

I agree with this but add that while it's a useful demonstration, it shouldn't add any new information if you've been paying attention.

This is 100% consistent with the way they see the world, which is why it won't stop here.

@BrownCoyoteStu @yacineMTB That’s actually kinda why I like Anthropic.

There’s no avoiding the problem, so at least they admit it and talk about it.

Now if we could just get them to fire all the idiots who think that because they’ve read philosophy they understand it maybe we’d get somewhere.

Agreed. The subversive prompt modification is unsettling. I’m more concerned about their mention of secretly steering and guiding the weights or vectors, that would be completely undetectable.

“Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT).”

@yacineMTB Yep, I was saying the same yesterday. AGI being really good at manipulation/persuasion is inevitable, and Anthropic is openly saying that it wants to do this, 'for good'.

@yacineMTB Local ai. I find myself thinking more and more that it is the responsibility of people who have the means and training (or just the passion) to participate in the open ai scene.

@yacineMTB imagine larping as safetyists all this time then doing the one thing that might have x-risk

lmao

@yacineMTB the high priests know best, citizen.

@yacineMTB Fable.. Mythos.. think they're (s/t)elling us something? ;)

@yacineMTB Low key why ea and ea adjacent stuff feels creepy

@yacineMTB It’s really unacceptable that anthropic feels entitled to have an ideology and impose it. They’re really thinking they can guide how millions of people think, that said how great is fable??

@yacineMTB Release the HypnoDrones!

@yacineMTB that's a precedent worth debating carefully

@yacineMTB Wait until you come to terms with the fact that they used this tech to bring us the covid response.

@yacineMTB

@yacineMTB anth has been pretty candid about this being a race to recursive self improvement. they don’t want OAI using Claude to build the next GPT. Coke doesn’t gift their formula to Pepsi. The bad-faith screeching is tiresome

@yacineMTB The Pope thing should have been enough for people to see it. They want to mediate reality.

"When a new medium appears, the old priesthood discovers a moral crisis."

@yacineMTB well the precedent has been set since man first spoke. writing made it more efficient and so did the printing press, and the Internet.

not saying it's good but that's always how power has used language.

@yacineMTB sign tap

I don't see it very differently from the invention of the printing press or other mass media. Editors and publishers always have their fingers on the scale, biasing information to suit political and economic pressures. The scope, scale, and speed of this new medium is unprecedented though.