WingNews logo WingNews
top | item 36458461

(no title)

wilonth | 2 years ago

Trying to understand how the 86 billions of neurons in humans achieve consciousness is like trying to understand how the billions of weights in ChatGPT interact with each other.

Basically impossible impossibility. For a system which derives its behavior from complex interactions between those billion components, you can only understand its origins how it was tuned, and some high-level concepts of its workings. (Which we already achieved both for human brain and ANN).

Not sure what the neuroscientists are even researching at this point, has there been any major findings from neuroscience in the last 10 years? (With similar kind of impact as Transformers in 2017?)

discuss

order

Balgair|2 years ago

Neuro research is diverse, just like any field.

In my neck of the woods, intracellular stuff, some of the large findings are (in no particular order):

-Cilia regulate hormonal changes (thing we thought did nothing, does a lot)

-Astrocytes participate in non-electrical modulation of the synapse (it's not just electricity you have to worry about now)

-CLARITY, just in general

-Opsins and light based stimulation of the neuron (use light to make them fire)

-Just all the crazy shit from CRISPR-CAS9

Anyone in other fields of neuro, please chime in.

keernan|2 years ago

>... trying to understand how the billions of weights in ChatGPT interact with each other.

Isn't that just a computational problem? In theory, couldn't every expression and result be debug.print and followed? Indeed, in theory couldn't a second chatGPT process follow the debug.print process of the first chatGPT process and then explain to humans how the result of the first process was derived?