wool bee pondering her wool orb

。o O ( this is not a flower)

bumble drills hole into flower it is otherwise too fat to enter

bumble drills hole into flower it is otherwise too fat to enter

。o O ( this is not a flower)

wool bee pondering her wool orb

understanding if 2 images are the same image or different images is surprisingly complicated

encountered some jewelry bees. very beautiful but also brood parasitic. your stance towards them will ultimately be determined by your stance towards macropis oil-collecting bees

1. A bit of evolutionary biology. I'm really intrigued by a new perspective piece from Steve Frank that explores connections between how natural selection creates systems that generalize and recent work in machine learning about the surprising capabilities of massively overparameterized systems.

most evolutionary novelty is regulatory. you can see this nicely in arthropodes, mostly shared genomes achieve high morphological diversity by tuning the activity of genes. the same gene expressed throughout all of development achieves different results than the same gene expressed only briefly

whenever i see a crabspider the kill bill music plays in my head and i wonder if i should let nature take its course or smite the spider for the sake of the bees with which i have a covenant

bumblebee emerges from flower, butt first, then gracefully takes off

i saw some bumblebees engage in nectar robbing. they chew holes into the sides of flowers for which they are too fat to fit in. smaller bees which can not chew holes themselves may stick their tongues into the existing ones made by carpenter or bumblebees

2 bumblebees quarreling over a flower

This book, like christianity the first 3000 years, has a gigantic scope and i wanted to collect things i found interesting: sodomy should probably be an insult for people who are rude to their guests, not for anal sex

everything about bees seems optimized to amuse me, which begs the question, how where these traits selected before i was born?

bluesky does not let me upload videos anymore for some reason, so unfortunatly you are all being starved of cute bee flower interactions

so i think fable might be completely useless to me. my memory file is full of deep learning and bio, but just small CNNs and fruit fly development. with memory deleted i can at least start a convo with fable, but it switches to opus if i just ask something like how many bee species exist in my area

i think these safety filters might be bugged, if only they had a really competent coding agent to do something about it :/

A cute beetle vomited on me in self defense

i dont think you co-evolved with this plant you little idiots

claude correctly predicted how long it would take for my potato peeled thumbnail to grow back. gemini overestimated by 3 weeks

humans are an excellent source of salt

recently a source from deep within bee cognition revealed to me that urban bees perform worse on cognitive tests than non-urban bees, unpublished though. i came across this earlier paper which maybe suggests the opposite? urban bees have larger brains royalsocietypublishing.org/rsbl/article...

why is biology difficult? do we have all the types but are missing the laws of their behavior? are we missing some types? do we have all the types but do not know all possible instances? do we know types, laws and instances but can not know their states? why cant we know their states?

i am curious how continuous learning will be solved. it is often said humans are more efficient learners since they need fewer training examples than DNNs. but this is only true if you do not consider evolution to be part of the learning process

i like cognition but i dislike consciousness and i hate, hate, hate, hate everyone who constantly forces me to think about it. NYAAAAAA FUCK THIS SHIT NYAAAAA

during training an NNs weights take on a shape that encodes symbols necessary for a task. animal brains do the same, but the genome also encodes a developmental program that leads to a brain with pre-initialized weights encoding some concepts, a bee knows what a flower is before it ever saw one??

having a tender moment with a sweat bee

The interaction they are destroying is mostly human viscera with car hoods

aftermath of violent altercation between canada goose and egyptian goose in, you guessed it, western europe. The canadian attacked the slightly smaller egyptian unprovoked. i witnessed something similar recently, canada goose attacking crow. sadly for me they never attack humans

"ouch! i think there is a chance that it thinks im prey" www.youtube.com/watch?v=Cl0g...

Lay not you hand upon me simpleton

why dont my stupid models understand that they are supposed to do what pleases me and not min/max their loss when the going gets rough?

life update: i trained a model 3 times on bugged data before noticing. source of the bug: me. but at least i got bee-themed grippy socks now

parabel of the digging bees. Me: adaptive, general intelligence, applying knowledge competently. You: pattern matcher, pudding for brains, lack of critical reflection

i feel like i should spell out which animals im comfortable embodying, that is if i found out a coworker drew me as one of those animals in a compromising situation i would not be disturbed. among insects: dipterans but only brachycera the true flies, all hymenoptera, all lepidoptera, all mantids...

what the hell is her problem?

im re-reading gravitys rainbow and it occurs to me all great novels are about the synthesis of sex drive and death drive. however i have only read 2 authors so far and i should get at least a third data point in

Arent Genomic sequence transformers kind of strange? they are probably strange in similar ways to LLMs, but LLMs operate in natural language for which we have an intuition, but for genomes we dont. Maybe the more transparent interpretability work on LLMs can be translated to genomics

very strange but also very interesting moment with some lab mates who have no familarity with deep learning. i dont need to understand anything about their problem, i only need to change the number of output channels for the new data set and in 2 days their task will be fully automated

hah, someone experiencing the mirror shock that I experience when I think of anything biological: "how does any of this even *work in principe*, this is clearly too poorly factored and incoherent to function"

i decided to write the code for my current project mostly by myself, with occasional input from claude. it does mean things move at 0.1% of their possible speed, but honestly what would i even do with my 17 hours of daily waking time? just wander around the gardens all day? thats no way to live

apparently the friendship between achilles and patroclus meme doesnt depict an ancient statue, but a still from a surprisingly artsy claymation short. it certainly does depict a very friendly relationship between the 2 www.youtube.com/watch?v=4zhE...

i think the impressionist, stream-of-conciousness depictions of events and environments in gravitys rainbow could work well as a cartoon. translating it into exaggerated character designs instead of narration. slothrops chapters are like looney toons episodes anyway

Our paper is on the cover of @currentbiology.bsky.social : a strawberry poison frog with the most common color morph, living up to its common name 🍓. In this issue, we uncover the genetic basis of color variation in frogs from Bocas del Toro, where blue, green, yellow, orange, and red morphs occur

what secrets are down there?

Its common name suggests lowly status, but the least flycatcher belongs to the genus empidonax which means mosquito lord and it belongs to the family tyrannidae, literally tyrants. what to make of this?

How finegrained is the for you recommendation algo? if im getting a bunch of posts about insects, including many about caddisflies and i click 'show less of this' on every post about caddisflies does the algo know i specifically dont want caddisfly content? or will it now penalize all insect content