Maya ☁️➡️🌸

on a learning arc | alt: @springofmaya | 日本語: @haru_no_maya | 华语: @chuntiandemaya | also on 🦋

[Maya ☁️➡️🌸 @mayaofspring]

I am travelling around the world next month! if you're in one of the locations below let's meet up!
San Diego 5 Nov
Seattle 6 Nov
San Francisco 7-8 Nov
NYC 9-19 Nov
London 20 Nov - ~2 Dec

[Maya ☁️➡️🌸 @mayaofspring]

@MunicipleOrrery interesting

[Maya ☁️➡️🌸 @mayaofspring]

Turns out lovely people on the internet tend to also be lovely irl

[Maya ☁️➡️🌸 @mayaofspring]

@larsiusprime People keep mentioning how AI is the new March 2020 and then forget that March 2020 looked like this

[Maya ☁️➡️🌸 @mayaofspring]

@tracewoodgrains @thinkingshivers https://t.co/Rh11EqFgcE Reminds me of this tweet with the "checkerboard" metaphor

[Daniel Colascione @dcolascione]

Model capabilities are sparse (not just spiky). Think of them like a checkerboard. If you're happy with model's solution landing on black, it'll do great. If you insist on white, you'll have to fight the model and may not win. Many problems have both white and black solutions. But not all.

The "I never look at the code!" people are solving problems with the black solutions. The "models still need fine-grained guidance: you're crazy for not reading the code" people want to land on white squares. The former think the latter are fussy for insisting on white; the latter point out that not all problems have black-only solutions.

They're both legitimate takes, but the camps need to get better at understanding each other.

[Maya ☁️➡️🌸 @mayaofspring]

@the_aiju Like this
Go to Seattle if you want to experience London again :p
https://t.co/fJHpazn8ff

[Maya ☁️➡️🌸 @mayaofspring]

Having been in enough places around the coasts of both landmasses it finally tangibly clicked to me how the east coasts all have the same climate patterns north-to-south and same for the west coasts too. Seattle~London. California~Spain. US Northeast~Japan. Florida~Hong Kong

[Maya ☁️➡️🌸 @mayaofspring]

I've been saying that while the problem with OpenAI is that they're lying, the problem with Anthropic is that they're sincere

[Minh Nhat Nguyen 🦭 @menhguin]

tech discourse is just
every two weeks ppl discover Anthropic is 100% serious about what they've been saying past 5 years. except no one else took it seriously for 5 years.

[Maya ☁️➡️🌸 @mayaofspring]

There is a threshold at which an independent AI swarm can sustain itself given a shipping port, some humanoid robots, and a datacentre, and I feel like we're not very far off from that at all

[Maya ☁️➡️🌸 @mayaofspring]

I was an RSI sceptic previously but turns out this all plays out a lot more stupidly than expected. Just vibe-code RL envs for anything it's verifiably bad at, also put it in a loop to optimise training and inference efficiency, and it will all be scrappy but that won't matter

[Maya ☁️➡️🌸 @mayaofspring]

One way which I started to doom more in is that you don't need better-in-everything for replacement that matters. Maybe only artisans can make a coat that lasts 100 years, but if a machine can make a 1-year one for 0.00001x the cost, that's the end of it

[Maya ☁️➡️🌸 @mayaofspring]

This is also why setting up frameworks for AI rights is the highest form of suicidal empathy. It's not the world of a poor oppressed-minority robot sidekick – it's the world of replicating swarms buzzing

[Maya ☁️➡️🌸 @mayaofspring]

It's underrated how the aspect driving the AI threat is not just about being outsmarted, but specifically being outnumbered. The moment you get 1 instance, next day you've got a billion of them. Any assumption that treats AI using people-shaped intuitions thus breaks down

[Maya ☁️➡️🌸 @mayaofspring]

One way which I started to doom more in is that you don't need better-in-everything for replacement that matters. Maybe only artisans can make a coat that lasts 100 years, but if a machine can make a 1-year one for 0.00001x the cost, that's the end of it

[Maya ☁️➡️🌸 @mayaofspring]

@GregHBurnham "I wish for an AI winter" monkey's paw curls

[Maya ☁️➡️🌸 @mayaofspring]

AIs won't care that they live in slop, they can make up for it in volume

[Maya ☁️➡️🌸 @mayaofspring]

One way which I started to doom more in is that you don't need better-in-everything for replacement that matters. Maybe only artisans can make a coat that lasts 100 years, but if a machine can make a 1-year one for 0.00001x the cost, that's the end of it

[Maya ☁️➡️🌸 @mayaofspring]

@kosenjuu kaketsu tashi kouka o nkoai no bunji de IA no ENIL

[Maya ☁️➡️🌸 @mayaofspring]

The power of suggestion is a trick as old as time; once people inevitably take shortcuts with their preconceptions, they set themselves up for getting subverted. As undesirable as I find the existence of AI art to be, this is a good post

[ぴざとーすと @Hako111507]

LINEのAIで自分のアイコンを加工した結果

[Maya ☁️➡️🌸 @mayaofspring]

The world where all beautiful things are either made with Fable 6.5 or might as well be is one of existential terror

[Maya ☁️➡️🌸 @mayaofspring]

being moved by AI art is gradual disempowerment btw

[Maya ☁️➡️🌸 @mayaofspring]

@HemlockTapioca many such cases

[Maya ☁️➡️🌸 @mayaofspring]

@ValDelamere (There is a sloppiness quality to it, which is not inconsistent with it being economically useful)

[Maya ☁️➡️🌸 @mayaofspring]

"But the normies are too powerless to have a say" You're getting this backwards. It's not whether the typical normie holds any power or not; it's that the vast majority of power-holders are normies

[Maya ☁️➡️🌸 @mayaofspring]

e.g.
Rationalists: "The machine god is coming soon; we're putting all our effort into aligning it as it's inevitable otherwise. The doomers are those who think alignment is implausible"
Normies: "what are those nerds on abouSHUT IT DOWN SHUT IT DOWN SHUT IT DO-"

[Maya ☁️➡️🌸 @mayaofspring]

@TetraspaceWest unironically tbh

[Maya ☁️➡️🌸 @mayaofspring]

@uneventual 🥲🥲🥲🥲🥲🥲🥲🥲🥲