Hi Matt,
Excellent, I think we can progress a bit.
The columns are accurate. Let me give some more detail on the
examples. Let's just focus on flight for now. This is how the science
reality maps to the diagram for Wilbur and Orville Wright:
1) they have a goal of replicating flight, and they must acquire a
level of scientific knowledge that does not yet exist. They are
inspired by birds as a known case of real natural flight. Their
mission is artificial flight, but in the process they'll create
abstract formalisms that capture scientific knowledge.
2) Their particular priority is replication. Other science contexts
prioritise the abstract formalisms. This is the science/engineering
balance.
3) On the right, there is numerical, paper, hand-computed exploration
of the Bernoulli equation. There is now and has always been this
computer exploration of abstract symbols, mostly tedious arithmetic,
.... Especially iterative solutions to differential equations. All by
hand on paper, for centuries... Millenia...
All that happened in the mid 20th century was that the 'computer'
job/process was automated with electronic computers. My Aunt Pat, who
recently died, was such a computer in an era where human computers
went extinct.
This is the RIGHT science activity that is the explorations of the
prediction an abstract model (theory/model) to find out how
predictive the model is of the observed nature in the MIDDLE. The
construction of the model can originate in human thought, or it can
literally result from empirical work done in MIDDLE and on LEFT.
When theory RIGHT, and observation done on, MIDDLE/LEFT match, job
done. Publish.
3) inspired by bird (MIDDLE) wings, they built wooden/canvas wing
profiles and 'replicated natural lift physics' in a wind tunnel. This
is empirical work of the LEFT kind. It captured the lift part, not the
propulsion part, of flight the way birds do it. The essential physics
of bird flight was conserved in the wind tunnel, not the bird's way of
getting it (feathers etc).
---------------
Empirical work need not always demand literal replication of MIDDLE
physics, but if it does then as a task it is a distinct category of
work and fits on the LEFT.
Example.
In LIGO, a massive experimental apparatus was built to detect MIDDLE
naturally produced gravitational waves.
There was no replication of MIDDLE gravitational waves. If there was,
then the scientists would engineer 2 black holes and fire them at each
other, thereby replicating nature... On the LEFT.
Example: CERN and the Higgs boson.
Unlike gravitational waves, the proof of a prediction of Higgs boson a
made by theory (the standard model of particle physics, RIGHT) and
this required human- engineered Higgs bosons. Just like everywhere
else in science, computing their way around the model is not
scientific proof that Higgs Bosons exist as the model predicted.
This time, the MIDDLE natural Higgs boson is missing. Not impossible,
just hard to find. All the empirical work happened on the LEFT.
So they built a machine that behaved in a way they thought would make
Higgs bosons. This is on the LEFT.
Example: Fire (combustion)
This is a case where replication came first, before science even
existed! 500,000 years later, Lavoisier replicated combustion LEFT and
displaced an existing RIGHT theory (phlogiston), replacing it with
oxidation. There was no serious numerical exploration (RIGHT) work
then. But nowadays, if you create a combustion simulator, to help
design a furnace... There's a computer on the RIGHT exploring
combustion formalisms, not actual combustion.
---------
BTW, the only reason I put a real computer substrate on the RIGHT in
every case, is because computers have automated the manual exploration
out of existence. Nobody does it by hand any more. Sorry I didn't
explain this more clearly. If it has confused anyone...sorry!
------- moving on .......
Every science context I can think of fits the LEFT/RIGHT/MIDDLE model.
They are very distinct categories and cannot be confused with each other.
In no science of a natural phenomenon (MIDDLE), ever, has the RIGHT
been accepted as or observed to be an instance of MIDDLE.
So if there is even a remote possibility of a level of real LEFT/RIGHT
equivalence, it is so unique it must be explored in a very thorough
way so that this amazing *exception* in science is understood.
That means understanding science and being able to articulate that
understanding in a manner and at a level previously unnecessary.
-------
The pertinent and special fact of the framework when it comes to
neuroscience of the natural brain/mind (MIDDLE), and it's involvement
with the creation of an artificial equivalent (LEFT) for the purposes
of the science is the real possibility of some level of complete
equivalence between LEFT and RIGHT - a possibility that is unique in
science and unproved.
Here, in AGI, we therefore encounter that real science possibility. We
all need to understand the conduct of science well beyond levels found
elsewhere. That is the responsibility we all have. That is part of why
I am here.
-----
Have I expressed this clearly? Please choose any science you know of
and map it into the framework... See if you can find instructive test
cases.
E.G.....The framework is, guess what? ... A scientific model of the
conduct of science by humans. The model itself is an example of an
empirically determined theory! RIGHT. The natural scientists (MIDDLE)
were observed and the model and the observations match ...and guess
what? Find any counterexample, and the model will need work, and I'll
be happy to see it?
And.. ...there's more .....
Q. The LEFT activity, in this self-referential application of the
model to science itself .. is what ?
You tell me. I'm serious.
:-)
Colin
The next test case is special, too.... 'computer science'. But that
can wait.
On Thu., 4 Jul. 2019, 10:50 am Matt Mahoney, <mattmahone...@gmail.com
<mailto:mattmahone...@gmail.com>> wrote:
Colin, I think I know why nobody is getting what you are trying to
say. It's because your diagram with the left, middle, right
columns is wrong. Instead of (fire, burner, computer chip), (bird,
airplane, computer chip), etc. there are only the first two
columns. We developed burners and airplanes without using any
computer simulation. So the correct analogy for AGI is (brain,
computer) in the left and middle columns.
Natural phenomena like fire and birds showed us one way to achieve
the desired technology. But we also have electric heaters and hot
air balloons which do not resemble anything in nature.
What is your position on AGI research using this analogy?
On Wed, Jul 3, 2019, 6:42 PM Colin Hales <col.ha...@gmail.com
<mailto:col.ha...@gmail.com>> wrote:
Hi,
On Thu., 4 Jul. 2019, 4:17 am Berick Cook,
<berickc...@gmail.com <mailto:berickc...@gmail.com>> wrote:
"assuming RIGHT is the only way to AGI, no matter how long
it takes, and never knowing for sure it that is even
possible or what computer resources and model-kinds are
mandated if it was, ....is the current approach
exceptionlessly adopted, without a discussion of the kind
we are having."
I wouldn't say that it is exceptionless. Many people are
very interested in recreating the brain in some hardware
form.
Ok. Please cite it. I've literally been looking non stop, full
time, for 20 years for any proposal by neuroscience /
engineers of the kind that made neuromorphic chips, directed
at AI. If you do find it, please let me know.
Many believe that your LEFT is the only path to AGI.
However, we just don't understand the physics of the brain
well enough to recreate it yet. There is a lot more
neuroscience that needs to be done first.
I have done one part of it (the neuroscience). Others have
also. US and europe. It's done to a level that can be used, or
that can be used to continue it, but only if you do it. Most
of the useful results happened around 2010 in mine and others.
We understand enough to get started. But if you propose it
you'll get exactly what you get here.
To perform empirical science, you first have to have
something to perform your experiments on. There is no LEFT
system yet, and the empirical tests that have been
performed on existing RIGHT systems don't come close to
achieving equivalent results to the empirical tests that
have been performed on MIDDLE. But those tests on RIGHT
systems are being performed. The empirical science is
happening, and RIGHT is steadily progressing forward based
on the results of those tests.
The work on the (e)RIGHT is not empirical science. It's
abstract models only. The empirical work happens only in the
middle. The work on the right might best be called
'experimental theoretical science', .... Like computing the
Higgs boson in the standard model is, compared to actually
making a Higgs boson (Middle) or observing a naturally
occurring one(left).
I'm trying hard to get the correct categories of activity
applied to what gets done in AGI. Exploring models with
computers is 100% theoretical. Ensuring the results match
empirical results does not turn (e)RIGHT into empirical work.
It makes it good neuroscience.
AGI is purely theoretical science at this point, because
it doesn't yet exist in any form. If you believe that LEFT
is the path to AGI, that's perfectly reasonable. If
someone else believes that it can be done with RIGHT,
that's perfectly reasonable too. But it is completely
unreasonable to say that everyone needs to abandon RIGHT
and focus only on LEFT, or that because RIGHT hasn't yet
achieved MIDDLE it is impossible for RIGHT to do so.
I never said it's the only way! I said that the right way of
finding out how to get to AGI is to do the *complete science,
like everywhere else.* This means using (e)LEFT as well, and
competing the science activities as found everywhere else in
science. It did not say you 'can't' on the left. I said that
whether you can, on the left, is discovered by doing both left
and right. Until then you don't know.
The reason I am here doing this is because (e)left of the
inorganic kind, never gets discussed or even posed as part of
neuroscience.
Dorian Air has posed an engineered natural tissue version for
(e)LEFT for AGI. That's it.
When the original silicon replacement experiment happened,
what the replacement did was replace the neurones with a
silicon model ((e)LEFT, omputer). It did not create a silicon
neuron.
I look forward to your citations, so I can check them. I have
scoured libraries and the world's literature looking for any
of it. Perhaps you have it! I'd love to see even a sign of
anyone making an AGI artificial brain proposal that hasn't
been a computed model neuromorphic computers are not that
proposal, btw.
---:
I've redone the silicon replacement thought experiment. After
editing and checking, I'll post the result.
Cheers
Colin
On Tue, Jul 2, 2019 at 8:08 PM Colin Hales
<col.ha...@gmail.com <mailto:col.ha...@gmail.com>> wrote:
BTW, happy to keep the discussion going
.... and many thanks to those that have engaged so
far. You've actually been training me to defend my
position and you've been doing a good job. And I hope
I am being receptive enough and informative and a
respectful way. It's not always easy when the
frustration meter red-lines.
:-)
cheers
colin
On Wed, Jul 3, 2019 at 2:10 PM Colin Hales
<col.ha...@gmail.com <mailto:col.ha...@gmail.com>> wrote:
Hi Steve,
The light is flickering on? I hope so. :-)
I am located in Melbourne Australia (not
Florida!). I am literally going back to Melbourne
University solely to spend the rest of my
so-called 'career' getting this matter ironed out
once and for all. Facilitating the formation of
the corrected science of AGI has taken the place
of actually building the AGI. I have to end this
strange era. It is mission critical for AGI and
too important to cut loose. I never dreamed my AGI
idea would end beached on this strange cultural
sandbar. But there it is.
I have thought many times about doing a
professional-level video mini-series (cartoon/live
combo) aimed at teaching the material I am
delivering to this thread (and that I want to
publish to the science literature). Meanwhile I
want to deliver the whole text of the chip design
concept to the thread as a first-pass at creating
the material for the paper.
I start back at Unimelb in a few weeks. I'll be
asking around for facilities to create videos and
I'll have a better grip on a script.
Thoughts?
Colin
On Wed, Jul 3, 2019 at 12:31 PM Steve Richfield
<steve.richfi...@gmail.com
<mailto:steve.richfi...@gmail.com>> wrote:
Colin,
I think I see the communications problem here.
This subject is a bit too complicated to
present piecemeal on a forum like this.
I suggest breaking it down in several videos.
It would be best to present this to a
skeptical expert.
Where are you located?
Steve
On Tue, Jul 2, 2019, 6:49 PM Colin Hales
<col.ha...@gmail.com
<mailto:col.ha...@gmail.com>> wrote:
Hi Steve,
I'll try and commenting in-line, if that's OK.
On Wed, Jul 3, 2019 at 7:54 AM Steve
Richfield <steve.richfi...@gmail.com
<mailto:steve.richfi...@gmail.com>> wrote:
Matt (and Colin),
I would like to clarify what I think
is your real question:
Nothing complex works the first time.
You build it, you turn it on, it
doesn't work, and you start debugging it.
It is very close to impossible to
debug anything without a solid
understanding of how it should work,
though I wrote a book about how to go
about such things, entitled "Advanced
Logical Methods" available free in the
Library of FixLowBodyTemp.com
Anyway, there has been a variety of
neuromorphic chips made over the
years, and there was no unknown
physics with these. Have any of them
been made to work satisfactorily? What
would be the debug plan with Colin's chip?
Just to be clear, neuromorphic chips
implement of model of brain physics that
literally is an electrical equivalent
circuit that uses non-brain
electromagnetism within which (if the
model is accurate) you can measure
voltages qualitatively similar to those
found in the brain. They can be quite
predictive of those voltages. When used in
a neuroscience, they sit under (e)RIGHT as
a computer, and when the voltages in the
brain match the voltages prodcued by the
model, it's exactly the same process as a
digital computer when the digital number
matches nature. It is not a replication of
the physics that produced the voltages in
the brain.
I have specified what "DEBUGGING" looks
like here, and I described it in detail
under the previous post: TESTING FOR
GENERAL INTELLIGENCE IN A NORMALISED
SCIENCE FRAMEWORK (e)
image.png
Your goal is an exploration of two
different kinds of potential AGI
(equivalent to testing the substrate
independence hypothesis) You have a test
and a control. Double-blind. You
compare/contrast two different brains in a
learning context and see when models of
brain physics (e) RIGHT tracks (or not)
the original physics (e)LEFT in a learning
context. You can start with a single
'cell' and build up and compare with
(e)MIDDLE.
The word DEBUGGING is a misleading word,
in context. This is empirical science.
There are two kinds of scientist involved
only one of them is 'programming a
computer'. One build the (e)RIGHT brain
(model-based, in this case a neuromorphic
chip) and (e)LEFT brain (brain-physics
based, any physics, not just my preference
but anyone exploring any possibly
essential brain physics). You are
exploring similarities and differences,
not 'BUGs'. The word BUG assumes you
already have an expectation of what is
right. In this context you do not know.
You are exploring to let nature inform you.
If this were the CERN collider, (e)LEFT is
the collider, (e)RIGHT is the standard
particle model.
The only strangeness here is that we have
the unique privilege of testing for the
only place in science where (e)LEFT and
(e)RIGHT might literally be indistinguishable.
The GREAT barrier I see are "field
effects" other than those Colin has
considered. My own favorite is the
variable conductivity within a neural
tube. If a tube abruptly goes from
conductive to insulator at it's inner
circumference then the Hall effect
would be minimal. However, if there is
a region of increasing resistance at
the outside of the conductive region,
then there would be a major inhibiting
effect from even tiny nearby magnetic
fields.
Remember, fields drop off as inverse
linear rather than inverse square from
long wire radiators, so interneuronal
field communication is MUCH easier
than it might first appear.
I suspect there are lots of such
things, and they don't all scale the
same as size changes.
Further, some things might be VERY
sensitive, like our own sodium levels
where we can become VERY sick if they
are a little "off".
I suspect Colin is trying to skip some
important steps between where head is
at now and new hardware - like simulation.
The normalised science framework /does
accommodate simulation/: Indeed it is
mandated. That is literally (e)RIGHT. The
compter exploration of a model of nature.
I have never tried to skip anything. What
I am trying to do get get the science to
/stop skipping the empirical work
(e)LEFT/. And my choice or anyone else's
choice of the physics in (e)LEFT is
irrelevant to the argument. It's the
process of doing the science to find out
which physics is essential and which
isn't. Anyone can dig out their pet
physics like 'microtubule quantum boson
X', and if such a thing is
empricially proved essential under
(e)LEFT, in the above testing context, by
literally putting that identical quantum
mechanics on the chip, then and only then
do you get to know something about what an
AGI can be made of. You get to compare
this with (e)RIGHT, a computed model of
the physics including "microtubule quantum
boson X". Yes it might even involve sloppy
electrolytes. It may even entail
engineered natural tissue. Anything goes
.....
It's the fundamental categorization of the
activity in science that is being
recognised here, not a specific preference
for a design of an artificial brain.
/The framework does not skip anything/.
It's the reverse, the science currently
skips all (e)LEFT activity. There is no
evidence anywhere of anyone even proposing
it, let alone doing it. This is a provable
fact of the science conduct.
This repair to the science conduct is 100%
agnostic to anyone's proposal for the
conserved brain physics in (e)LEFT. The
figure above it created to accommodate my
preferences. It is simply the normal
science framework correctly presented with
my proposition positioned accurately
within it (e)LEFT, and the existing
approaches also accurately placed in
(e)RIGHT, just like everywhere else in
science.
Maybe there is a highschool science
fair level demo and early debug
prototype - like large models of
several neurons in a fish tank full of
salt water - in effect an electrolytic
computer like I mentioned in an
earlier posting?
Anyway, I see Colin's field theory
concept, but Colin needs to see a lot
more, including a clear path through
design and debugging, before we can
really discuss this.
So, Colin, having read this, it is
back to Matt's question, like how do
you hope and expect this to work?
Thanks both of you for your
persistence here.
Steve
I already delivered exactly this days ago.
You have it all.
I still seem to be unable to get the
fundamental message across: that AGI can
and will only arise and be proved when its
science is properly populated with all the
activities normally associated with
science (as depicted in the LEFT, MIDDLE &
RIGHT above). If this one science
ultimately gets to completely abandon
brain physics (and all empirical work
involving it), then the practitioners must
empirically prove this as a possibility,
not assume it is possible. You have to
deal with the brain physics first. It's
the only proved natural general
intelligence exemplar. If this was
'flight', then the exemplar would be the
natural bird. The brain is our bird. You
examine the physics of the bird for the
origins of flight because its your proved
natural case.
*Brutal fact: The science of AGI to date
has been conducted under (e)RIGHT by a
community that assumes the truth of a
hypothesis that is only proved by doing
something that community never does, has
never done, and doesn't (apparently) have
any awareness that it is normal science
practice everywhere else in science. This
whole thing is a cultural blind-spot, a
historical accident. Everybody in it is
wearing these glasses and seems to be
unable to take the glasses off. *
*
*
In the above, I'm showing you the
CERN-collider experiment of AGI. If you
really want AGI and whatever dreams are
fulfilled by it ....those dreams currently
have been 100% bet on an assumption that
isn't proved, whose empirical proof isn't
even thought about, and that constitutes a
unique, unjustified lack of empirical
science. Only here in AGI. Nowhere else in
science. The AGI enterprise is critically
dependent on getting the science right,
and throwing $Billions at a project based
on an assumption never proved or even
tested properly. And if that assumption
proved true, it would be /unique in
science. A proof that demands maximal
empirical scientific acuity, not none./ I
am trying to help you all understand that
and how incredibly, hypercritically
important it is to the fate of the AGI
enterprise. And I am doing it while
watching everyone in AGI inhabit a
blizzard of ongoing evidence predicable,
stereotypical, industrialised
under-performance whose explanation lies
in the science I am saying has not been done!
and boy, is it hard work being the
messenger bearing the news! I hope I am
getting closer. :-) Patience to death and
beyond!
Colin
--
Berick Cook
Independent Developer
AI Research <http://airis-ai.com/>
Games / Software <http://berickcook.itch.io>
YouTube Channel <https://www.youtube.com/c/berickcook>
*Artificial General Intelligence List
<https://agi.topicbox.com/latest>* / AGI / see discussions
<https://agi.topicbox.com/groups/agi> + participants
<https://agi.topicbox.com/groups/agi/members> + delivery options
<https://agi.topicbox.com/groups/agi/subscription> Permalink
<https://agi.topicbox.com/groups/agi/T87761d322a3126b1-M67455e0a13ab6512798d2b14>
