[ExI] all we are is just llms
Ben Zaiboc
ben at zaiboc.net
Tue Apr 25 08:02:59 UTC 2023
On 25/04/2023 00:00, spike wrote:
>
> In any case, I don’t know what that spike-trains business is about.
> Ben, do explain please. Sounds kinda cool.
>
Try:
http://www.neuwritewest.org/blog/2015/1/3/ask-a-neuroscientist-whats-a-spike-train
Gives some information, but doesn't explain how they arise, and why they
are digital signals.
Not to get too technical, neurons 'fire' by depolarising the cell
membrane. This means that the normal distribution of electrical charge
across the membrane is reversed. This is all to do with ions of various
types and ion channels (proteins that span the membrane and allow
specific ions through) and pumps (that actively pump certain ions
through the membrane).
In total, the membrane is like a capacitor that tends to settle at a
certain voltage. Certain events can make the membrane quickly
'discharge', reversing polarity, to a second specific voltage. I forget
the exact numbers now, but they're easily found by searchiong the web
(or asking an AI). A single depolarisation is a 'spike'. You could
represent this like so: ________|_______
That's a section of neuronal membrane at resting potential being
depolarised to a specific voltage then quickly resetting to normal. This
happens very quickly, and the voltages are always the same. So,
effectively, this is a digital signal: 000000000100000000.
Spike trains are clusters of these signals that travel down the axons
(pipes that extend out from each neuron, carrying the signals away, to
other neurons) and dendrites (similar pipes that bring signals to the
neuron body).
The integration and general signal processing that the neuron does is
complex and I don't understand it all, but can be thought of as
something like a logic gate processing incoming digital signals and
producing an output. Each neuron has many (thousands, usually) of
incoming dendrites, and (usually) a single axon leading away to other
neurons.
That might have been in more detail than you wanted or needed, but is
still a very dumbed-down, 'without numbers' version (and the numbers are
important to fully understand the process), and there are many
variations on this, so if anyone knows neurphysiology, please don't be
outraged at this version. You need to study this stuff in a lot more
depth to get a proper idea of how it works.
Also, I am not an expert. This is my graduate-level-and-below
understanding, and is probably out of date now. But the broad principle
is sound.
One other thing I will say: biology is messy. These 'digital' signals
are not nice and neat like bytes and words in a computer. Each
individual spike is the same as all the others, but the 'trains' I keep
mentioning are of different lengths, occur at different frequencies, and
of course occur in millions of individual axons. This is why when people
say "The brain is not a computer!!1!" I groan. It's just not that
simple. The brain is like a computer in some ways, and not like one in
others, is really all you can sensibly say.
Ben
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230425/4ee78dd3/attachment.htm>
More information about the extropy-chat
mailing list