tailcalled's profile picture. ❤️ @aliceryhl ❤️. Programmer, psychometrics nerd, hobbyist gender researcher. Autistic.

Tailcalled (@tailcalled.bsky.social)

@tailcalled

❤️ @aliceryhl ❤️. Programmer, psychometrics nerd, hobbyist gender researcher. Autistic.

Pinned

I'm generally not providing subtle hints for places that would be beneficial for others to investigate. When I tweet about something, it might be a half-baked take, or it might be a conclusion from a more thorough investigation. I don't know what *you* should do.


Tailcalled (@tailcalled.bsky.social) reposted
horsetwting's tweet image.

Tailcalled (@tailcalled.bsky.social) reposted

remember big data?


tailcalled's tweet image.

No, I wouldn't say I'm a normie. I'm an *aspiring* normie.

SomeBrashAtom's tweet image. No, I wouldn't say I'm a normie. I'm an *aspiring* normie.


The pizza principle: the more the constituents of a food resemble those of pizza, the better the food tastes.


Curious too

Has anyone had success stories (or can point me to writeups) with fine tuning frontier LLMs?



Tailcalled (@tailcalled.bsky.social) reposted

biologists continue to mistake single-cell technical sampling noise for meaningful cellular heterogeneity. the apparent sparsity of single-cell data is due largely to technical artifacts like transcript capture and sequencing depth--if there are hundreds of thousands of…

"Only ~0.02%-3.1% of [a cell's] genome" is being transcribed at any given moment. Other interesting takeaways from this new paper: > If you pool together a bunch of cells of the SAME type (like primary immune cells from a mouse's spleen), and you measure the transcription for…

NikoMcCarty's tweet image. "Only ~0.02%-3.1% of [a cell's] genome" is being transcribed at any given moment.

Other interesting takeaways from this new paper:

> If you pool together a bunch of cells of the SAME type (like primary immune cells from a mouse's spleen), and you measure the transcription for…
NikoMcCarty's tweet image. "Only ~0.02%-3.1% of [a cell's] genome" is being transcribed at any given moment.

Other interesting takeaways from this new paper:

> If you pool together a bunch of cells of the SAME type (like primary immune cells from a mouse's spleen), and you measure the transcription for…


Tailcalled (@tailcalled.bsky.social) reposted

"Only ~0.02%-3.1% of [a cell's] genome" is being transcribed at any given moment. Other interesting takeaways from this new paper: > If you pool together a bunch of cells of the SAME type (like primary immune cells from a mouse's spleen), and you measure the transcription for…

NikoMcCarty's tweet image. "Only ~0.02%-3.1% of [a cell's] genome" is being transcribed at any given moment.

Other interesting takeaways from this new paper:

> If you pool together a bunch of cells of the SAME type (like primary immune cells from a mouse's spleen), and you measure the transcription for…
NikoMcCarty's tweet image. "Only ~0.02%-3.1% of [a cell's] genome" is being transcribed at any given moment.

Other interesting takeaways from this new paper:

> If you pool together a bunch of cells of the SAME type (like primary immune cells from a mouse's spleen), and you measure the transcription for…

Tailcalled (@tailcalled.bsky.social) reposted

evopsych: hunter gatherers today depend on trade w industrialized societies. which implies our hunter gatherer ancestors did too. bc industrialization is recent, those trading partners must have been time travellers. placing humanity in a causality violating time loop. troubling


Tailcalled (@tailcalled.bsky.social) reposted

Toaster, microwave, cauldron all give nice results!


Is there any other emoji it gets equally confused about? I tried a few but couldn't find any.

gpt5 gets so fucking confused

arm1st1ce's tweet image. gpt5 gets so fucking confused


Tailcalled (@tailcalled.bsky.social) reposted

there's half dozen studies with two digit sample sizes and mixed replications. The most cited study finds differences in preference for a pink pot, which it attributes to biology. No difference in preference for a stuffed toy, according to the study that is a gender neutral toy.


Tailcalled (@tailcalled.bsky.social) reposted

so far I assumed that the evopsych factoid "monkeys show the gendered toy preferences of children" was from a robust literature showing dymorphism in preferences for humanoid/animate toys being extrapolated by ignorant people. I decided to look up. I was so wrong. It's so bad.

This post is unavailable.

Tailcalled (@tailcalled.bsky.social) reposted
miniapeur's tweet image.

I guess the big open question is whether it can be fine-tuned with e.g. robotics data to be more accurate.

This looks visually nice but seems too imprecise to be used as a world-model...



This looks visually nice but seems too imprecise to be used as a world-model...

Actions are persistent in the world.



Tailcalled (@tailcalled.bsky.social) reposted

hot take: getting negatively polarized is deeply embarrassing, not a normal psychological phenomenon, and not something to be proud of. it is a midwit behavior that should be shamed out of any circle where people claim to care about seeking truth


Tailcalled (@tailcalled.bsky.social) reposted

16. suckerpinch / tom7. a sigbovik legend. software engineering as postmodern art. weirdly only 190k subscribers

adrusi's tweet image. 16. suckerpinch / tom7. a sigbovik legend. software engineering as postmodern art. weirdly only 190k subscribers

If self-driving cars are more reliable than human-driven cars but stuck in regulatory hell, why isn't Elon Musk constantly complaining about that?


Loading...

Something went wrong.


Something went wrong.