Mike Tintner wrote:
Colin:The part of my idea that freaks everyone out is that there is no
programming involved. You can adjust the firmware settings for certain
intrinsic properties of the dynamics of the EM fields. But none of
these things correspond in any direct way to 'knowledge' or
intelligence. The chips (will) do what brain material does, but
without all the bio-overheads.
If I remember right, you're embodying the objection of a comment
there, which said basically: "but the brain is just firmware and data
- there is no software." Well,. this dynamic approach is potentially
fine and exciting. But as exposition it suffers from the lack of an
example. How will your machine first a) perceive objects in the world,
(even if it's an artificial world), then b) associate responses to
those objects, and c) choose between alternative responses - all
*without* software? (The basics of intelligence). I'm not asking for
the whole theory, just your v. basic idea of how your firmware
approach will begin to produce an intelligent response without prior
guidance. Without that, people clearly won't "get the idea" as you
suggest. (De Bono offers some such idea somewhere - The Mechanism of
Mind, I think).
P.S. I don't believe your thesis that a "cane toad killer" has its
"learning capability" switched off. That's the sort of thing I expect
from ethologists who treat animal behaviours as if they're completely
wired in. But I would think any roboticist would appreciate that
that's impossible - no matter how restricted the repertoire of an
animal, if it's living in the real world, it will continually confront
variations on its problem-solving - cane toads say in unusual places
and positions - that demand a fresh adaptive response and learning.
The point is very subtle. This is how product deployment occurs.
I have an 'empty /but supercharged/ AGI' at my 'factory'.
I hook it to a 'cane toad killer' body.
We get the basic physiology of motions sorted.
Then....I teach it to be a cane toad killer.
In it's supercharged state it'll /also be able to learn maths and really
bad tennis!/
But I don't teach it that.
I put it through 'cane toad killer boot camp'.
When it can cope with all manner of field (as in Australian bush) -based
novelty associated with being a cane toad killer able to sniff out cane
toads .... then I clamp down on those areas of brain dynamics that mean
it can rapidly acquire new knowledge in other areas . All existing
knowledge, the dynamics of being a cane toad killer - these will
remain. Nothing you can do to the robot will enable further learning.
This will be mission-fatal for some cane toad killers who, say, fail to
recognise very novel threats. Just like real cane toads don;t recognise
the threat of the motor car tyres. I can teach the CTK to avoid traffic.
This 'nobbling' is quite plausible in the chip architecture. It's what
happens in nervous systems. You can't just 'flip s switch' to turn on
the advanced learning. The delivered hardware itself will be constructed
to be a cane toad killer and nothing else. I download and 'burn' a
specific version/subset of the chip which is 100% cane toad killer and
is unable to become anything else for the same reason cane toads can't
learn maths.
Meanwhile I go on to the next project for the '/supercharged/ AGI'.
Say... the 'smart weed killer' that lives with farmers and knows exactly
what a weed looks like, eats it into a DNA-dead state and then shits it
into the farmer's soil. Release a small flock of these into a field...
no more chemical fertiliser. They can't reproduce. They return to base
if they are sick (or their associates carry them back).
We only need 1 'supercharged' AGI. It will have to cooperate with us to
create these little beasties.
You are perfectly right that existing robotics experts will say the kind
of learning is impossible. For existing robots that's true. But then I'm
not having anything to do with existing robots, am I?
Another favourite application for me "the incorruptable and relentlessly
honourable company". That it, a 'legal entity' incorprated literally to
inhabit the role of a company. It literally 'feels' the balance sheet
and revenue statements. It panics about cash flows. Feels extatic when
profit is good. No longer do we need 'rules of incorporation'. The
company literally IS the AGI. If the company "goes bad" you take it out
and shoot it. The process of giving birth to a real company is literally
giving birth to s specialist AGI - the actual company itself attends
board meetings... fun eh! The hard question - do you invite it to the
coporate dance night? He he.
cheers,
colin hales
-------------------------------------------
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/
Modify Your Subscription:
https://www.listbox.com/member/?member_id=8660244&id_secret=123753653-47f84b
Powered by Listbox: http://www.listbox.com