llm
according to media
a llm is a bunch of parameters in a layer
where each parameter connects to parameters in another layer
and there are many parameters in each layer
and there are many layers in a llm
but google search does not show web articles
that explain how many connections each parameter in a llm has
and a human brain is neurons connecting to each other
but probably not in layers
and each neuron connects to a lot of other neurons
so if a llm is made more brain like
then maybe it could become more intelligent
maybe by each parameter having more connections to other parameters
maybe by having clusters of parameters instead of layers of parameters
maybe by allowing llm to self organize it's clusters of parameters
example:
one cluster with many parameters
with few connections to each other
that generate one functionality
and another cluster of few parameters
with many connections to each other
that generate another functionality
and a third cluster of many parameters
with many connections to each other
that generate yet another functionality
example:
in a self driving car
one big cluster of parameters generate vision
and a smaller cluster of parameters generate hearing
and another cluster of parameters generate driving
Comments
Post a Comment