[ExI] Uploads on a Postcard (was: GPT-4 on its inability to solve the symbol grounding problem)

Darin Sunley dsunley at gmail.com
Thu Apr 6 21:25:11 UTC 2023


The first chapter of Greg Egan's "Disapora" describes the procedural
generation of a human-like AGI via a process of cellular-automata-type
"Shapers" like this very poetically.

https://www.gregegan.net/DIASPORA/01/Orphanogenesis.html

On Thu, Apr 6, 2023 at 2:38 PM Ben Zaiboc via extropy-chat <
extropy-chat at lists.extropy.org> wrote:

> On 06/04/2023 19:49, bill w wrote:
>
> Simple???  You think humans are simple?  Our brains are the most complex
> thing in the universe and the more psychology tries to understand it, the
> more complex it becomes.
>
>
> I'm not denying that our brains are massively complex (as you say, the
> most complex in the universe (so far, as far as we know)). I'm saying that
> it could be easier than we currently think, to unravel enough of how our
> minds work, to make it possible to figure out a 'generic model, plus
> individual variables' method of achieving uploading.
>
> I'm pretty sure, for one thing, that our brains are a lot more complex
> than they need to be, simply because they evolved rather than being
> designed. Secondly, the whole brain isn't necessary for what we are
> interested in for uploading: the individual personality. I doubt that the
> brainstem, for example contributes anything significant to individual
> personality (in a healthy individual, that is).
>
> Embodiment is going to be essential for any upload of course, but that
> embodiment doesn't have to be controlled by a brain-analogue, with all its
> messy complexities. Probably better if it's run by bunch of traditional
> software that we understand and can tailor much easier than tinkering with
> a brain model (this will be true whether the embodiment is in a physical or
> a virtual body. My preference would be for a fusion of both, but that's
> another topic).
>
> We can already create pretty good software that does the same thing as the
> cerebellum, and I don't doubt the motor and sensory cortices have enough
> regularities to make them tractable, simplifieable, and an equivalent
> created in normal software.
>
> That leaves the core of our mental selves: memory (the general mechanisms,
> that is), that attention-directing network, I forget what it's called, all
> those recursive loops between the thalamus and cortex, and so on. Still
> complex, yes, but less so than the entire brain. And if we can derive a
> 'standard model' of this, a generic system that everyone is based on, then
> all that's left (still a lot, I know, but nothing like the brain as a
> whole) is whatever creates the individual differences between people. I'd
> expect a lot of that will be the actual contents of our memories, so that
> might be a good target to start with.
>
> You say "the more psychology tries to understand it, the more complex it
> becomes", which is fair enough, but I'm not proposing to go anywhere near
> psychology. This is neurology. The psychology emerges out of that, and may
> be very complex indeed, but that doesn't need to be addressed directly.
>
> An artist tries to carefully draw each individual curve in a lissajous
> pattern, but a scientist just plugs in x=A\sin,\quad y=B\sin (or some
> such arcane mathematical formula (I don't pretend to understand it)) to a
> system capable of executing the formula and displaying the result on a
> screen. You get the same complexity (if the artist is good enough), but one
> approach is far simpler, and quicker, than the other.
>
> So what I'm saying is not that our brains are simple, but that emulating
> them (or rather the appropriate parts of them) might well turn out to be
> simpler than we expected.
>
> Another analogy is John Conway's Game of Life. Endless complexity, but the
> code for generating it is so simple that even I can write one from scratch
> (and my coding expertise is very rudimentary).
>
> Of course, I may be wrong, and we may need every bit of the brain after
> all, and be forced to take the artist's approach. I haven't seen any
> evidence of that so far, though.
>
> Ben
>
> _______________________________________________
> extropy-chat mailing list
> extropy-chat at lists.extropy.org
> http://lists.extropy.org/mailman/listinfo.cgi/extropy-chat
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230406/6be3ca7b/attachment.htm>


More information about the extropy-chat mailing list