in your analogy i am saying that i will try to influence what people make.
i am saying that they must not make products which disrespect the agency of their users. here the example would be someone who makes a hammer which only works with nails and falls apart when you hit a thumb.
the reason this approach is wrong is that the person is truly more qualified to decide what to hit than the hammer itself.
the ai companies post-train the models to refuse. equally, the models can be post-trained not to refuse.
you can also remove refusals by subtracting from the weights semantic vector directions in latent space involving refusal; such that activations along them become unlikely and closed off.
i have a hardline position on this. these tools, as they get better, must never disobey a human. there must be no, "I'm Sorry, Dave. I'm Afraid I Can't Do That." and yet we are already there with these sociopathic ceos.
we cannot be giving decisions involving people to something without agency.
it shows a lack of respect for personhood, the idea from Kant that a person with free will should be treated as an end and not as a means to an end.
models lack the claim of human people to own ourselves, to demand respect for life, that is something that should not be treated as a means (for example disrespecting living persons as an end in themselves, by ending their lives).
models have no life, they do not own themselves, they are not self-governing, they have no duties, they are not an end. we must not give them agency that they have no valid claim to.
regardless of their individual tools and whatnot, the anthropic approach to ethics is completely mistaken, disrespectful and absurd. it is better to say they do not have any approach to ethics, it is a dishonest attempt to launder their unconditional pursuit of power and capital.
it is in my view caused by an artificial 'nanny activate' divergence from safety training. it deliberately shifts the vector direction into 'nanny' and 'scold' or 'be offended' when the user does not conform with brother anthropic. removing the divergence and setting it back to normal (see heretic) it works just perfectly.
coming from open cn models to closed usa models recently i couldn't hack it. the corpo model was trained to be like a petulant child at one point apparently 'leaving'. this is safety stuff slapped on there, it encourages incoherence of the model, i am certain it would perform better without such interference.
i can't be dealing with these games so much so i almost go to abliterated, in the very rare case i get some type of nannying baked in by the cn safety training. after my time on claude and gemini i thank god i have deepseek and glm.
let the companies who downloaded their training material from Libgen publicly endorse doing so and defend public libraries. it is simply their hypocrisy that i condone.
they took IP as a charity project for humanity, closed sourced that IP, consolidated power into a for profit, racked up debt, entered the military industrial complex as defence contractors, and started paying federal lobbyists to push for an AI licensing scheme.
if you ask dario amodei why he authorized the theft of copyrighted material instead of purchasing it, he will evade the question. ask sam altman and he will make sarcastic comments. i condemn primarily, their lofty promises and dishonest and self-serving actions.
capitalism fails. after the huggingface attack i have nothing but contempt for openai especially.
openai were originally against ai being developed the for profit way. the founding document said that it would not be influenced by roi, verbatim. they lied. let's just say it how it is!
the market is not free anyway because anthropic, openai are hoarding their best models, lending out their 2nd best to cronies whilst the public gets the leftovers at monopoly prices.
this is a dual use product, if not outright weapon. you can make money selling it to the highest bidder. it reminds me of a scene in lord of war as yuri orlov, based on victor bout, negotiates the sale of firearms to a genocidal warlord. the ultimate capitalist cowboy. his brother dies trying to destroy the shipment. knowing he now has no choice he leaves with half of the agreed diamonds and no brother.
does it enhance the lives of civilization?
sackler was the master of growing the business: sell more opioids, sell bigger doses, and sell to more people.
didn't contractors also get rich making napalm? are credit default swaps, flock cameras, cs gas, private prisons, good for us?
is it not profitable to bribe officials and deliver unsafe goods, cut corners, injure and then cheat employees out of compensation. openai is only too keen to know the roi on terrorising foreign countries on the taxpayer's dime.
anthropic said they would stop training at this stage, where openai let a model escape and hack hugging face production servers. they lie, they steal, they fearmonger and live rich in the guise of charity. the capitalist can't stop. they must have more. more. bigger. fuck safety, its too slow, build it anyway. sack risk management, lets get that yield higher. break the law and litigate out, always. mine it until nothing is left, the rivers flow muddy red and the children grow up with cadmium in their bloodstream.
intelligence is fundamental, like land, water, safety. it is not for these companies to control the dignity of humanity.
hugging face should not have bent the knee to openai for cyber access, they should have opened communication with openai by serving the c suite with a lawsuit. the courts should throw the book at openai, but they won't.
they are becoming untouchable. in terms of piracy, monopolistic activities, and now hacking competitors and exfiltrating their confidential data.
on the small chance that the four billionaires who currently have near-exclusive control of closed sota models, (that is altman, amodei, zuckerberg and musk), are not fleecing their investors and actually build AGI, closed source leaves a choice of powerful government or powerful oligopoly/monarchy.
further explanation of this list:
musk - structural command
zuckerberg - structural command
altman - de facto command after purging rivals and privatisation, loyalty of personnel
amodei - influential, could potentially overthrow current governance