hckrnws

For a split second there I believed there is a new Distill publication! Their articles were the most inspirational and eye-opening resource on my beginnings of ML journey, the quality of visualizations definitely made lasting impact on my mental models of _what is going on_.

My favourite goes definitely to The Building Blocks of Interpretability (https://distill.pub/2018/building-blocks/), those images landed in a lot of my university presentations and the dog made everyone immediately interested ;-)

Really cool, and although I am not quite up to date on biology research (relevant here), I believe there were exciting results showing mechanisms like electric fields acting as signals for cells to differentiate in some species, i.e. "telling a cell where it is", and supposedly guiding "what it should do/become", etc.. I believe this was a result from studying Axolotls, the amazing self-regenerating (and severely threatened of extinction in nature) salamanders[1].

Side note: if we needed more reasons to conserve the amazing and enormous spectrum of life, one more reason is this kind of discovery that might enable better understanding (and maybe enhancement one day) of cell growth and regeneration in humans. Also showing that biology in many ways is extremely far ahead of what humans can achieve with current technology or will for the foreseeable future (as much as the automata example is very neat, it's nowhere near self-assembling full working and self-reproducing creatures from a compact genetic code!).

It seems you can donate directly to help Axolotl conservation (which again is critically endangered), seems really important if you can help! [2] (although there are of course many other means to help if you're interested in conservation in general!)

[1] https://youtu.be/7cLaU_agj6k?&t=86

[2] https://www.moja.ong/programs/axolotl-habitat-conservation/ https://www.moja.ong/donar/

There is ongoing research on neural cellular automata, as they seem to be a very efficient way to generate pretraining tokens: https://arxiv.org/html/2603.10055v1

Very cool! So much potential for interesting stuff in this algorithm. I also recently made a minimalistic browser playable cellular automata [1] implementation for a game jam - I think it's relevant enough to link here, might be interesting to some.

[1] https://nmcela.itch.io/conways-voxels

- is that a blog that you are familiar with that teaches you neural networks in the most animated visual manner that absolutely blows your mind off?

- because it finally clicks and you realize how it works on the inside?

- https://visualrambling.space/neural-network/

- i found one but it doesn't have text content explaining the math, the reason behind gradient descent, the need for a partial derivative etc.

- I would super appreciate if someone has a resource for cnn, rnn, and other types of neural networks like this

Morphogenesis data compression is pretty impressive, the human genome is only ~700Mb of data.

This has been shared a few times here. It looks cool but I don't understand the usefulness of this tech. This is likely a demo to show some capability, what that is though I don't understand.