jorbs

5 hours until sts 2 *vibrating noisily*

a pack of kittens closes in a a large dog standing on a chair

David Attenborough [whispering]:

“As the predators close in on their helpless prey, we can only watch in silent horror, for it is not man’s role to interfere in the hierarchy of nature.”

oh no! rest in peace.

chapter 2 definitely my favorite. it's early enough in the game to stand out and embedded in the theme of the chapter in a beautiful way.

personal least fave was ch. 4 cause i didn't play smb3 so the reference missed for me and it just felt random. wouldn't say that that was its fault though xD.

insofar as "reasoning" means "a word for how humans do the things they do", llm's aren't ever going to do that, because they aren't ever going to be humans. insofar as it means can they compute things to succeed at tasks, sure. that isn't a new thing for technology to be able to do.

that is about whether or not llm's can perform some tasks as well as humans, but doesn't address whether what they do will ever have equivalence to what humans do. is an llm ever going to feel its anxiety soothe when it holds a newborn child, for example?

calculators and airplanes can perform certain tasks better than humans can. llm's are a tech hoping to perform some more tasks better than humans can, but they aren't ever going to be equivalent to human reasoning. they could surpass it in many ways, but they still wouldn't be the same thing.

sure, am familiar with the history of humans thinking they are different than other things due to something about how they think. the thing is that however we think is in fact the way we think, so it's sort of tautological if there's any important on being human, which most humans have.

also like, another angle, we've been doing whatever it is we do for thousands of years, and it has a tremendous number of nuances etc., so assuming you can successfully replicate it and swap it out with llm output because it works sometimes is extremely extremely fragile handling of sentient life.

for what it is worth it is not that hard to make a compelling ethical argument that humans should replace themselves with other things, like nothing, or ai. it's just not compelling to many humans, because we're humans.

it doesn't apply very relevantly to humans because humans are (literally) grandfathered in. we are our species. whatever it is we do that may or may not be reasoning is the thing which we, our species, does. arguing we can replace it with ai slides into asking why we don't replace ourselves with ai.

n.b. he is repulsive. i just maybe don't need my media to tell me that repeatedly, it could be more useful if my media just told me the things that were happening.

in a media landscape where everything about him has to be repulsive, i have no real ability to get like... information. it's not a great media landscape.

one weird thing about ■■■■■■ ■■■■■ is that everything about him is treated with approximately equal outrage on social media. like, him being old is treated with about as much scorn as him remodeling the white house, which is treated with about as much scorn as him raping children.

it understands that it needs the keys, but doesn't understand how to navigate to the superelite to get emerald consistently

starting killing iran's schoolchildren this early in the process is extremely extremely fast ideological colonization

(it has to hold 60% winrate for 500 runs to tick its assistance down one number. started at 50 free aoe damage 47.5k runs ago).

in terms of actual learning, it plateau'd for quite a while at 22 free damage per turn, which i think is where it starts stalling against donu deca if it doesn't understand it needs to play additional damage probably. breaking through that hump feels like maybe a big deal. down to 20 now.

spirebird, neural network for losing at slay the spire, progressively getting worse at losing.

i am having way too much fun with this. found a way to save ~200mb ram per instance so i'm up to 32 workers, and batching gpu updates (running them synchronously every ms instead of immediately every time a worker reports in) got me ~4x throughput as well.

i wonder if anyone watches donald █████ speeches for the same reason they watched alex honnold's taipei free solo

I guess our elected officials were right this past year. I should have stopped all this from happening

if this were vs. a sphinx i would answer "pluribus" and feel 70+% happy with my answer.

is this pluribus?

10 hour study with me at Columbia showing lunch

Her in the library

DHS illegally arrested me please help Text against image of her knees

Columbia student detained today is a day-in-the-life influencer with 100K followers. Her stories right now go from her studying for a Genetics exam in a nicely edited "ten hour study with me at Columbia" video, to a photo of her knees with the caption "dhs illegally arrested me please help"

the party that wants to require ID to vote abruptly invalidating a thousand people’s IDs overnight seems like a pretty giant flashing red light

it's unjustifiable to pull people out of their community and put them in a camp because they have expired paperwork. to do so would be barbaric and inhumane. not a relevant question, and look inward at whether you think the government should be able to put you in a camp for ~anything~.

when my roomba pauses to upload pictures of the interior of my house to whatever server it's sending them to i like to think that it's praying

i have been coding a bit to get through whatever cold/flu thing i have