Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

I extracted the safety filters from Apple Intelligence models

github.com

Post reply on HN
Print view View original

441–450 of 455 posts

Re: I extracted the safety filters from Apple Intelligence models

#441

Post by WillAdams »

grues-dinner wrote: post #23
WillAdams wrote: post #13

Earlier quoted context omitted.

Legal requirement to maintain a trademark.

In what way would (A|a)pple's own AI writing "imac" endanger the trademark? Is capitalisation even part of a word-based trademark? I'm more surprised they don't have a rule to do that rather grating s/the iPhone/iPhone/ transform (or maybe it's in a different file?).

Imagine how a court case would go where a lawyer wheeled in a Mac and noted it was a default OS install and projected onto a screen the results of querying Apple's AI about Apple products and it failed to adhere to the trademark guidelines.

Re: I extracted the safety filters from Apple Intelligence models

#442

Post by kube-system »

MisterTea wrote: post #427
kube-system wrote: post #388

Earlier quoted context omitted.

Which still does not demonstrate that they believe it has opinions. Natural language is how you interact with an LLM -- interactions will mimic human interaction, even for those who realize it is not sentient.

They were under the impression they could in fact change the AI's mind. So yes, they did believe it has an opinion. They believed it was sentient and able to think for itself. Do not underestimate peoples inability to distinguish between a very clever Markov chain and actual intelligence. The future is going to be ... interesting.

>They were under the impression they could in fact change the AI's mind.

They aren't really wrong here. LLMs are often trained on input. Have you considered you might just be taking their anthropomorphism a little too literally? People have used these anthropomorphic metaphors for computers since the Babbage machine.

Re: I extracted the safety filters from Apple Intelligence models

#443

Post by SV_BubbleTime »

avianlyric wrote: post #424
SV_BubbleTime wrote: post #177

Earlier quoted context omitted.

I would hope! But no one actually believes Google is politically neutral do they?

Evidence suggests they’re about as neutral as you could hope. It’s not like Google search is some kind special tool used only by the elite. It’s pretty trivial for political scientists to pump queries into Google and measure the results. Which is exactly what many have done. There’s been plenty of independent research into political bias of Google search results, and plenty of lawsuits that have gone fishing via disc…

Maybe you haven’t followed…

The CEO hosted a cry session about broken hearts and how they as a company would resist when Trump won in 2016.

The black nazis, female popes, etc. No, Google isn’t neutral.

Re: I extracted the safety filters from Apple Intelligence models

#444

Post by martin-t »

lazide wrote: post #439
martin-t wrote: post #438

Earlier quoted context omitted.

Not to sound like I am rejecting the possibility but can you tell me how you got that information? I would be very helpful for convincing people in general to have something more concrete to go on that a random comment.

I build those systems at a company that you definitely are aware of. I can’t discuss it further due to my NDA. Feel free to ignore that any of this exists of course - it makes our lives easier. It’s a constant arms race regardless.

Then I have 2 questions:

- Why are they not flagging more content? Am I right they're boiling the frog slowly? Do they lack an endgoal because management does not yet understand the power of these tools?

- Do you do your job poorly on purpose? Did you take it so somebody else wouldn't build an even better system? Did you think you could influence it in a direction which does not lead to total surveillance? (I assume any reasonable intelligent person would be against further increasing the power imbalance corporations have against individuals for both moral reasons and because they are individuals themselves who understand the machine can and will be used against them too.)

Re: I extracted the safety filters from Apple Intelligence models

#445

Post by spookie »

0points wrote: post #289
dkdbejwi383 wrote: post #267

Earlier quoted context omitted.

It has been suggested - although I am unsure if there is strong evidence - that the word "bear" is a euphemism along these lines, meaning "brown one" for the since-forgotten original name for the animal, as it was allegedly believed to be either too frightful to say aloud, or would summon a bear.

It's not suggested, the historic use of noa words is a fact. See https://en.wikipedia.org/wiki/Noa-name

Oh! Cape of Good Hope was renamed that way when portuguese mariners discovered they could go around the areas more susceptible to "freak waves".

This is a problem even today, some have said it is due to hotter currents coming from the Indian ocean meeting the cold Atlantic. But the judge is still out on that one.

Good documentary on rogue waves: https://www.youtube.com/watch?v=EfNc_6EjbMU

Re: I extracted the safety filters from Apple Intelligence models

#446

Post by xk_id »

energy123 wrote: post #210
userbinator wrote: post #59

China calls it "harmonious society", we call it "safety". Censorship by any other name would be just as effective for manipulating the thoughts of the populace. It's not often that you get to see stuff like this.

This is the rhetorical tactic of false equivalence. State censorship by an autocracy with the objective of population control is not the same thing as a private company inside a democracy censoring their product to avoid bad press and maintain goodwill for shareholders. If you want solid proof that it's not the same thing, see all the uncensored open weights models that you can freely download and use without fear of…

Except Apple has a country’s worth of users, whose livelihoods are reliant on them. The “state democracy” is right now more subordinated to tech oligarchies, than vice versa.

Re: I extracted the safety filters from Apple Intelligence models

#447

Post by Asraelite »

mystified5016 wrote: post #412
plasticchris wrote: post #320

Earlier quoted context omitted.

Which is a reflection of morality, of sorts.

Morality and law are completely disjoint. On a Venn diagram, it's two circles separated by about a lightyear or so.

What? They couldn't be more closely connected. In a democracy, laws are to a large extent a reflection of the wishes of the voters, and voters want what they believe is moral.

It's true there's no casual relation in the other direction, if that's what you mean - law does not define morality.

Re: I extracted the safety filters from Apple Intelligence models

#448

Post by blitzar »

qingcharles wrote: post #419
stripline wrote: post #337

Earlier quoted context omitted.

My kids watch streamers on YouTube and the common replacement is “frick”. It’s said so often that they started using it saying things like “what the frick!?” so I had to explain to them that’s essentially the same as using the real word.

I wonder why they didn't just use the "frack" from BSG?

Takes a person of culture to watch BSG

Re: I extracted the safety filters from Apple Intelligence models

#449

Post by immibis »

anton-c wrote: post #331
immibis wrote: post #318

Earlier quoted context omitted.

Do you still use swastikas as symbols of peace and love because you don't want white supremacists to define your language? I strongly doubt you do that. Whether you like it or not, the Nazis defined what the swastika means now.

It's still seen in the countries that used it that way and is seen as benign. It can be easily summoned with the Japanese keyboard. It's seen on Buddhist temples all over Asia.

Do Japanese people speak the same language as you and I?

Re: I extracted the safety filters from Apple Intelligence models

#450

Post by lazide »

martin-t wrote: post #444
lazide wrote: post #439

Earlier quoted context omitted.

I build those systems at a company that you definitely are aware of. I can’t discuss it further due to my NDA. Feel free to ignore that any of this exists of course - it makes our lives easier. It’s a constant arms race regardless.

Then I have 2 questions: - Why are they not flagging more content? Am I right they're boiling the frog slowly? Do they lack an endgoal because management does not yet understand the power of these tools? - Do you do your job poorly on purpose? Did you take it so somebody else wouldn't build an even better system? Did you think you could influence it in a direction which does not lead to total surveillance? (I assume…

Have you stopped beating your wife yet?

Cut the bullshit.

Post reply on HN