mero

KIM KITSURAGI - The lieutenant leans in confidentially. “You wouldn't be able to hear if he were wearing anal beads,” he whispers.

how about tomorrow 5 or 6pm my time for my character creation? :O

Nogizaka 41st single promo pic. Members’ costumes smoothly transition from a light indigo near the top to white on the bottom of the skirt. There’s floral details on their arms set on clear mesh like to look like they’re directly on their skin, kind of reminds me of figure skating costumes a bit. The skirts are very long and wedding dress-like.

ETHEREAL!!

ooooo!!! i don’t have anything friday afternoon thru to sunday night so it could be any time during then!! i’m just trying to sleep early so if we do it at my nighttime id like to wrap up by 9pm my time at latest

梅澤美波 卒業のお知らせ (Umezawa Minami Graduation Notice)

oh shit…

is korogatta kane so narase a ska song? i dont know enough about music genres to be confident dhdjdjd

i am not one of them lol

"São Carlos is the first city in South America in numbers of doctors per inhabitant, according to a survey done since 2006 by UFSCar. In all, there are 1,700 PhDs, which represents one for every 135 inhabitants. In Brazil, the ratio is one PhD per 5423 inhabitants."

wow, huh

I don't know why, but the judge is now citing the historical precedent of Charles I of England. This is an incredibly ominous sign for Yoon.

im sorry, im not sure i understand what you mean in your point 2

correction to the first tweet: “because for many that’s NOT an option”

like if we’re going to put time and effort and money and research on trying to improve education, “how do we help learners out of isolation so they can connect with their peers” seems more valid than “how do we make AI tools better for learners”

yes, i’m not saying it’s a “individual responsibility” kind of thing where individuals need to choose to connect with peers instead of using AI because for many that’s an option. but as a society saying “these people can just use AI, that’s good enough” is a cop out

but the fact that these risks exist to me makes it seem even less worth it. it’s a bad tool for the specific task at hand and on top of that there’s a bunch of dangerous risks and “side effects” and it just feels like trying to fit a square peg through a round whole for no reason

but that’s more about ways in which chatbots and such seem harmful /dangerous in general as they might eh addictive or reinforce people’s biases and delusions and the like, which is a separate matter from the “fundamental flaw”.

i also just worry in general about how anthropormorphized these AI tools can be? i guess with chatbots specifically. simulating the experience of talking with another person when you’re not, people relying on that and getting instant gratification from the bot instead of connecting with peers

like you can say humans are also not 100% reliable and there’s teachers out there teaching innacurate things but… there’s at least more mechanisms for auditing that and holding people responsible than there is for AI tools

so it’s a slot machine where you can’t guarantee that the results are reliable. an experienced person can at least use it as a shortcut (say, an experienced programmer using AI autocomplete to generate boilerplate they’ve done 1mil times) & if it hallucinates they’ll see it. a learner can’t do that

well we know it’s impossible to make an AI that doesn’t “hallucinate” because it’s a phenomenon inherent to how it works, right? a large language model generating text is predicting what words are likely to come next in response to a prompt but it doesn’t understand the content and context

so i 100% understand people similar to me being helped by AI and using AI as that “peer” to bounce ideas off of, but again, i just think generative AI is fundamentally flawed for that purpose and people settling for that and becoming more isolated from real human peers sounds bad to me

speaking for myself, i have TERRIBLE blank page paralysis / anxiety and having anything to jump off of helps a ton, especially if it’s a new and daunting thing i’m trying to learn. and one thing that helps me HUGELY is having a peer to bounce ideas off of and work on a first draft together

and i’d prefer to work on making non-AI-based learning more accessible than on figuring out or promoting the “best ways to use AI for learning”

maybe other kinds of “AI” could be applied meaningfully to education, but “generative AI” to me feels too inherently flawed for that purpose to be worth it

i don’t doubt that there’s people managing to do real learning using AI, but also feel like the risks/harms outweigh the benefits, and that if people are having to rely on AI to learn that’s more indicative better ways of learning being too inaccessible?

like. that would make it way easier to identify the noise / “false positives” and throw them out. give me better search and i’ll take it, but i don’t need the computer to sound like a human talking to me, that literally just makes the output *worse*

the computer is more efficient at searching through preposterous amounts of data but it’s way less efficient (or incapable) of understanding and interpreting the content in context. so can’t it do the search but then just point me to what it found without “generating” the dumb unreliable summary?

but the more open ended and subjective a problem/task is, like precisely in that first draft blank page stage, the more it makes way more sense for humans to do that part. if the point is for technology to help ppl rather than replace them, it should be optimized to do the things humans are worse at

i think it’s also that it’s being optimized for the things computers are the worst at? the more open ended a problem is the less efficient a computer is at solving it. i don’t mind sucking at doing arithmetic by head and letting a calculator do that work for me bc the calculator does it better

i disagree with OP that all the thinking happens in the blank-page state but even if i AM doing proper thinking on top of an AI generated first draft, if im a beginner who doesn’t already know better i wont know if the AI gave me garbage, and then its “garbage in / garbage out”

like i would rather have an example of another person’s solution/thought process on something rather than having to figure it out from first principles but the AI isn’t a person and doesn’t have a thought process and i can’t trust it didn’t just give me bullshit…

i feel like i’m the kind of person for whom worked examples are extremely helpful, but the problem with doing that on top of an AI first draft is having no confidence on the quality level of what it spit out…