eestileib (she/hers)

@dalias@hachyderm.io @austinkocher@mastodon.social I've had ACLU people canvass me and I say "I don't think our values align". One guy just visibly sagged and said "it's the nazi thing isn't it?" Yeah, yeah it is.

@zzt@mas.to The post from Anthropic about how badass their interface was, and it was a giant pile of kludge because they hadn't heard of ncurses ...

@junebug@slime.global I suggest you switch to the distro I use, which replaces your impenetrable problem with a new impenetrable problem.

@root42@chaos.social @kenshirriff@oldbytes.space Ok but the ad does at least talk about how it's replacing your engineers' time spent doing repetitive calculations. LLM pitches are about how you can reduce you and your employees to drudges by removing the need for you to think.

@mayintoronto@beige.party @skinnylatte@hachyderm.io As someone who walked among tech people, one of the most frustrating things is that people Do Not Understand how little these people care about other humans, how destructive they are.

@JeffGrigg@mastodon.social @Catvalente@wandering.shop It's a billion degrees and they're all deeply afraid of cities.

@overholt@glammr.us Indeed, that quotation you pasted above infuriates me. It's the same energy as saying "why would you ever want to go see a statue when you can just look at pictures of it?" It's the same energy as "yes we killed your grandma but look she's in datacenter heaven and we have a chatbot of her. Same same, right?"

@thisislieven@c.im @RosannaSibora@fosstodon.org The "leaders" are doddering dotards.

@ricci@discuss.systems This is what ladder climbing at a mega corp did to me.

@mike_bowler@hachyderm.io The person in the bottom right corner is the only person who knows what the hell they're talking about though!

@monkeyninja@10base2.dev This is my conclusion as well. If what you're trying to do has been done before, chatbots will give you a degraded copy of it. If you're trying to do something new at some point you will have to think. A lot of people are content not to try to do something new.

@cube_drone@mastodon.social @golgaloth@writing.exchange Just as soon as I start shitting billiard balls, I will install this toilet.

@Benhm3@mastodon.social @amydiehl@mstdn.social AI flatters with absolutely zero shame or exhaustion.

@davidgerard@circumstances.run I know a person who wants to get railed when I see one (cause I often am one) and I'm looking at Pete Hegseth.

@Mr_GHARice@mastodon.social @matoakit@beach.city @maxleibman@beige.party There's a limit to wealth beyond which it breaks your mind

@amydiehl@mstdn.social But that would mean that Elon Musk might be a narcissist

@bob_zim@infosec.exchange @briankrebs@infosec.exchange That's my question as well, unless someone not from a company driven by religious devotion cosigns, I'm taking this very skeptically.

@fight@mastodon.fightforthefuture.org I mean when the scene is led by a bunch of weirdos like Sam Altman and Peter Thiel, yeah you're gonna get surveillance.

@darthnull@infosec.exchange @mttaggart@infosec.exchange @cR0w@infosec.exchange Your effort was not in vain

@GroupNebula563@mastodon.social @redsad@ohai.social Uncanny, they predicted Sam Altman.

@cinebox@masto.hackers.town @jonny@neuromatch.social

A malicious dataset abused two code-execution paths in our dataset processing (a remote-code dataset loader and a template-injection in a dataset configuration) to run code on a processing worker.

Am I reading this correctly as "our dataset processing is to prepend some text with processing instructions and an SGML tag, and then hand that blob to an llm" ?

@laprice@beige.party @ricci@discuss.systems Stanford is a moral and ethical vacuum so this is sadly not much of a surprise. They ran the Stanford Prison Experiment, after all, which is probably one of the major reasons why IRBs exist in the first place.

@spawnofzap@mastodon.social @GroupNebula563@mastodon.social It's Marx laughing that a capitalist is appealing to the labor theory of value to get other capitalists to buy his shit.

@DeltaWye@tiggi.es @jerry@infosec.exchange Seriously. I'm switching banks because the first one I created an account with only offers their app running on stock android. Like, screw you, why does Google get to decide whether I get access to my money?

@theloopfarm@mastodon.social @gwynnion@mastodon.social Planning a General Strike without knowing how to feed your strikers and their families is planning to fail. No coincidence that so many places in the US are making public peer food support illegal.

@chillybot@infosec.exchange Absolutely this. It's like the opposite of the _mens rea_ doctrine, where acts engaged in with complete disregard for any ethics whatever become especially permissible

@jxwojtowicz@mastodon.social @josephcox@infosec.exchange The 2016 election was 1933, this is 1940.

@VeroniqueB99@mastodon.social @gdinwiddie@mastodon.social It's a different thing. It's not chili, it's not spaghetti, it's not a hot dog, it's Skyline. If you have a preconception you will be disappointed.

@VeroniqueB99@mastodon.social Mierdas Touch

@quixoticgeek@social.v.st Mark Zuckerberg wrote and paid for all those laws.

@scott@sfba.social @danielcornell@mastodon.social They want to be the new Kennedys.

@wannabemystiker@expressional.social @quinn@social.circl.lu I think this is humanity, it's just that the cults in charge of the US are on track to be the worst ever.

@sabik@rants.au @lymphomation@mastodon.social @david_chisnall@infosec.exchange @tante@tldr.nettime.org "But Science!" is just the new talking point for AI astroturf. Notice how all the boosters started saying that and "jagged" at the same time?

@dagb@snabelen.no @arstechnica@mastodon.social Good choice Ars. 👍🏻

@courtney@pdx.social I've been wrestling with this myself, I pack snack bags to give away and want to minimize my waste, cause these are going to people on the street and probably half of the bags will get dropped where they are. So I'm using plain paper bags, but they need to be closed enough that they won't dump their contents. I've had kids so I don't trust folds or rolls for a morning, let alone passing thru multiple hands. I was using trombone clips, but ended up switching to staples because they hang on to a folded paper bag better and are far less metal per. I could use toothpicks or something to pin the top shut, but these are going to people who are often high at the moment they get them, so I'm hesitant about using stuff that could leave splinters. I had been using trombone clips and binder clips for so long that I had to go out and get a stapler when I decided that was the best option.

@geekysteven@beige.party He'd love the idea of the Library of Congress but be disappointed by the execution. I think he'd piss himself over bulldozers and radar guided counterbattery fire, but would be totally lost when it came to how to use/defend against drones.

@theloopfarm@mastodon.social @gwynnion@mastodon.social Sikh Gurdwaras are excellent models to look to here; they have a religious obligation to feed those who come to hear about their religion, and many extend it to food service for the larger community. More importantly, (particularly in India) they have a system set up to feed a lot of people that does not require endless amounts of single-use plastic. The fact that it's part of a religious practice also makes it more inconvenient for cops to raid the site, pollute the food, beat up the people eating, etc (like they did to the Black Panthers etc).

@VeroniqueB99@mastodon.social Obama was a decade ago. Time to move on.

@cmsdengl@mas.to @spaf@mstdn.social It's called a _Damnatio Memoriae_

@petealexharris@mastodon.scot @maxleibman@beige.party Yup this is it.

@mattsheffield@mastodon.social Second and third paragraphs, I'm totally with you. I guess where our perspectives differ here is that I don't think of llms as turning "meaning" into speech, they continue an existing fragment of speech based on probability. There's obviously a ton of probabilistic stuff our brains do to predict the next word, frustrating that expectation is how a one-liner works, for example. I think there are experiments you could run by asking "what's the next thing Henny Youngman is gonna say here?" and watching which neurons fire, in my mind that would more directly address how much human brains can be seen to function like llms. I think from where I'm standing that the study was more about how humans create utterances from their own nonverbal/conceptual internal model of reality, which seems like a fundamentally different kind of thing than what llms do. I think maybe we're seeing different things cause we're looking in different directions from the paper.

@VeroniqueB99@mastodon.social They were always part of it

@mattsheffield@mastodon.social This is interesting, thanks for sharing. There are definitely observations in there that are worth hearing about. I felt like the article just cliff-asserted "this means brains form statements using the same computational method llms do", and clearly I'm missing a very important step. From the info available in the article, "llms and brains are converging on the same computation model" feels much more like a faith based directional statement, or something inserted by an editor to drag clicks, than a thing that an experiment was set up to test. The article says that there's a substrate of neurons that seem to carry a kind of shared nonverbal model of reality, and that expressing a thought in a particular language seems to involve translating that through selected neurons that are closely-situated to the neurons activated by whatever state the model of reality is in. That's cool! And as far as I know, it's nothing like an LLM, which operates on a language based, syntactic base model.

@gullok@mastodon.social @joho@mastodon.online Live in the fucking USA for a couple of months, understand what it means to have a repressive government, then think about it again.

@adhdeanasl@beige.party Yeah I made it about four minutes into the new Infowars and thought "why am I doing this? this hurts to watch"

@cmsdengl@mas.to @spaf@mstdn.social Wow, I had no idea that term was a retroactive description, thanks that's cool to learn!

@mattsheffield@mastodon.social I'm unfamiliar with what "geometrically encoded" means for llms. Is what they're saying that "the weights of graph edges in an llm are analogous to a distance metric between synapses in the brain"?

@mattsheffield@mastodon.social I'm personally fascinated by how multilingual people's minds decide how to decide what language to use. My base languages are English and Portuguese (smattering of others particularly French and German), and I frequently notice that when I hear Russian my mind spends a few confused seconds trying to parse it as Portuguese. Or how when I'm speaking to a family member and they just use the Portuguese word in the middle of a sentence, if I've recently spoken Portuguese or Spanish, it's smooth, but if I've only heard English for a month, there's a bump while my mind figures it out. It feels like these guys are starting to explore the space where we can start looking at those neural correlates too, and that will also be interesting.

@rallias@hax.social @Binder@petrous.vislae.town I can lend you some if you need.

@sebastian@mastodon.cc @david_chisnall@infosec.exchange @tante@tldr.nettime.org I think the odds of that are zero, I'm assuming that Google will find someone to bribe before long.