Twitter | Search | |
Leonid Boytsov
C elegans has less than 500 neural cells, but it has basic sensory system and muscle control! It can reproduce and mate. With just 500 freaking neurons.
Reply Retweet Like More
Leonid Boytsov Jul 26
Replying to @srchvrs
a correction: the number of neurons is actually abt 300-430 depending on the sex, but there are also a few dozens other aux. nerve cells.
Reply Retweet Like
darya filippova Jul 26
Replying to @srchvrs
A nerdy note: even single cell organisms can reproduce 🤓
Reply Retweet Like
Leonid Boytsov Jul 26
Replying to @lynxoid
Great point, but I also commented that they can mate with each other (although most of them will self-reproduce). They do a lot of tasks each of which involves a lot of small decisions. 500 artificial neurons is likely not enough even for one small decision.
Reply Retweet Like
Leonid Boytsov Jul 26
Replying to @lynxoid
For example they move by controlling dozens of muscle cells. Good luck training such a thing by using RL with neural containing only 500 neurons. It would be many thousands to get decent results.
Reply Retweet Like
darya filippova Jul 26
Replying to @srchvrs
Ah, I see your point now
Reply Retweet Like
Grzegorz Chrupała Jul 26
Replying to @srchvrs
Many microorganisms can sense, move, and reproduce with no neurons at all.
Reply Retweet Like
Sunny'D Jul 27
Replying to @srchvrs @CryptoBari
Reply Retweet Like
Leonid Boytsov Jul 27
Replying to @gchrupala
Right even single cell is very complex, but this is a different level of complexity then voluntary propelling of a 1000-cell body. For a more compelling/obvious example: a fly sees, flies, mates, & does a lot of other things with just ~200k neurons. The smallest drone? :-)))
Reply Retweet Like
Crypto Bari Jul 27
Replying to @napstervab @srchvrs
Reply Retweet Like
Delip Rao Jul 27
Replying to @srchvrs @gchrupala
Neural networks are not the only networks in organisms. Pretty much all organisms, from plants to higher mammals, have intra- and extracellular networks that modulate functions and adapt.
Reply Retweet Like
Delip Rao Jul 27
Replying to @srchvrs @gchrupala
For eg, in humans, the role of endocrine networks and lymph networks is underhyped compared to the nervous system while they do affect massive changes in the functioning of the organism. Of course, evolutionarily all these different networks have co-evolved and co-adapted.
Reply Retweet Like
Grzegorz Chrupała Jul 27
Replying to @deliprao @srchvrs
Regarding the computational sophistication of single-celled organisms, I can recommend this book:
Reply Retweet Like
Leonid Boytsov Jul 27
Replying to @gchrupala @deliprao
Thank you for the reference if a single cell organism is much more complicated than I thought it only makes my point stronger
Reply Retweet Like
Fernando Pereira Jul 27
Yes, I loved how it addresses the evolution of multilayer control through internal reward signals
Reply Retweet Like
Leonid Boytsov Jul 27
I skimmed some parts: looks like a great read. Maybe someday :)
Reply Retweet Like