both those are related to information theory, but there are other things I legally can’t mention. shrug.

hahahaha fuck off with this. no, the horseshit you’re fetishizing doesn’t fix LLMs. here’s what quantization gets you:

• the LLM runs on shittier hardware
• the LLM works worse too
• that last one’s kinda bad when the technology already works like shit

anyway speaking of basic information theory:

but the research showing that 3 bits is as good as 64 is intuitive once you tie the original inspiration for some of the AI designs.

lol

not gonna give any money to a startup whose weird fucking founder thinks their primary product is AI rather than search, thanks

thank you for bravely rushing in and providing yet another counterexample to the “but nobody’s actually stupid enough to think they’re anything more than statistical language generators” talking point

lemmy-flavored redditors: uhm I came here expecting some nuance

also lemmy-flavored redditors: nuclear power plants opening is always good, why would I care that the companies doing it have a history of ignoring regulations, are treating nuclear engineering as loosely as software engineering, and are generating vast amounts of power just to chase an awful fucking fad and explicitly not to power houses or any worthwhile infrastructure? what do you mean read the article? why would I ever do that?

no, the article isn’t clickbait if it accurately paraphrases a source in a tone you disagree with

no, it’s not ragebait if reading it made you angry and you wish it wasn’t true, but it very obviously is

good talk, thanks

though to be honest, the fact that you think this is local-only and only affects business accounts perfectly demonstrates how fucking dangerous Proton’s marketing and design around this feature is

An AI-driven nagbot will surely fix these systemic issues — Huffington specifically claims the bot will “address growing health inequities” — and revolutionize healthcare, because “behavior change can be a miracle drug, both for preventing disease and for optimizing the treatment of disease.”

fuck off Arianna. there’s so many life-altering, commonplace diseases that behavioral changes don’t do fuck all for, and meanwhile some of the most beneficial behavioral therapy you can do is fucking impossible to get cause insurance won’t cover it in any achievable form. all this horseshit does is, as usual, shift the blame for beyond inadequate healthcare from a thoroughly broken system to the people suffering, because they supposedly didn’t try hard enough to change their behavior and get better.

thanks to this article specifically, Ludic’s going on a podcast with Robert Evans:

I just agreed to go on a podcast on a whim, then a friend told me it is with the host of Behind the Bastards, and I spat tea everywhere.

My to-do list today said "go for piano class" and "prepare for rental inspection".

I am unprepared for this level of prime time.

it’s a fucking tragedy what techbros, fascists, and gurus have done to the perception of what a trip even is. there’s this hilariously toxic idea that a trip must have utility or else you’ve wasted your time and material; these people will gatekeep and control how you process what should be a highly personal experience, because that gives them a subtle but powerful amount of influence over what you’ll take back from your trip. the Silicon Valley/TESCREAL crowd have even ritualized bad set and setting so much that they don’t need a guru personally present in order to have a bad fucking time.

the damage these fucking fools have done is difficult to quantify, largely because psychedelic research is either banned or firmly in the hands of those very same fucking fools. it’s very important to them that you don’t use your temporarily jailbroken neocortex for your own purposes; that you never imagine a world where they don’t matter.

the voice for OpenAI’s Sky got taken offline because they tried to copy Scarlett Johansson’s voice for it (she voiced the AI in Her) after she repeatedly told them no (via this mastodon post which links the original NPR article):

oh come the fuck off it, OpenAI’s marketing presents sora as exactly a magic automate entire movie clip button. here’s OpenAI marketing the stupid thing as a world simulator which is fucking laughable if it can’t maintain even basic consistency. here’s an analysis of how disappointing sora actually is

tonight’s promptfans are fucking boring and I’m cranky from openai’s shitty sora page crashing my browser so I guess all you folks doing free marketing for Sam Altman can fuck off now

it’s very funny that OpenAI’s being so public with this, cause there’s a bunch of embarrassing shit in there:

• several admissions that OpenAI only ever used the veneer of being an open source non-profit as a recruitment, fundraising, and marketing tool — there was never a plan to function as a non-profit once they’d attracted attention
• “The core algorithms we use today have remained largely unchanged from the ~90s. Not only that, but any algorithmic advances published in a paper somewhere can be almost immediately re-implemented and incorporated. Conversely, algorithmic advances alone are inert without the scale to also make them scary.” aka, there’s no real breakthrough here, the fancy markov chain just looks scarier after you feed it the world’s data
• musk’s apparent plan was for OpenAI to implement Tesla’s self-driving tech. when they said no he threw a tantrum and decided Tesla could do this AGI thing itself. this one isn’t surprising at all but it’s very funny that musk assumed GPT was the key to making his self-driving shit work
• the general amount of cult jargon combined with clueless MBA chatter
• fucking slatestarcodex posts steering top-level decisions at OpenAI