Ben,

Here is a sad situation. Everything you say here is correct, but your logic
is fatally flawed, leading you onto a suicidal path.

First, note that there is a BIG difference between "brain imaging" and
"brain diagramming" In brain imaging, pictures of parts of the brain are
produced that hopefully convey some useful information. In brain
diagramming, a database containing the circuit diagram is produced while
the brain is pureed, with NO images of the physical structure being
produced, except internally along the way to producing the diagram.

On Fri, Jun 22, 2012 at 4:30 PM, Ben Goertzel <[email protected]> wrote:

>
> FWIW, when I discussed Steve's design for a new brain imaging instrument
> with some mind-uploading friendly brain imaging experts (who have designed
> new brain-matter-analysis instruments themselves), their reaction was that
> it seemed like the data would probably be too noisy to be useful.


Yes. At least initially the data WILL be too noisy to ever simulate what is
seen. However, the data as to how we are wired would be VASTLY superior to
anything we now have, albeit too flawed to directly simulate.

I believe there are techniques to overcome the noise, but you won't find
them in neuroscience. Once the math is fully understood, you can go back
and "fudge the terms" for correctness. The computations would then be
"correct" though somewhat different (noise) than the original neurons did -
but who cares, because the simulated neurons would probably function better
than the originals. However, this is all later. The initial results, noise
and all, will at least tell you what feeds what, which is NOT (yet)
determinable with present methods.

They weren't ready to call it a worthless idea, but they didn't feel it was
> as promising as various other areas of brain imaging research...
>

Of course this is a function of goals. Sure there are other cheaper
experiments to tell you things of present interest, but NO other approaches
to provide any sort of whole brain diagramming, albeit too noisy for direct
simulation. They are interested in curing diseases, while you are
interested in producing intelligent machines, and these goals require VERY
different research.

>
> Since it's not my area of expertise, when I got that opinion, I decided to
> ignore the matter and wait for the brain imaging experts to figure out that
> stuff...
>

This won't work, because their ONLY sources of funding are health-related.
Sure brain diagramming may have some value for health, but I can't make a
good case that it will help cure ANY particular disease. So long as ALL
brain research is funded through the Department of Health, there will be
little of value for AGI.

>
> I don't think diverting resources from AGI to brain imaging is a
> particularly good idea.


Neither do I. I completely agree with you on this point. All I was looking
to do was to spread the word as to what could be done, and hopefully drum
up enough interest to eventually find funding.


> Brain imaging is way better funded and staffed than AGI already.


Not any sort of imaging that is useful for AGI. Again, AGI needs
diagramming, not imaging.


> And other stuff is way better funded and staffed than either of these.


Too vague to understand what you are saying here.


> $$ and human effort for exploring alternate speculative brain imaging
> methods should be taken from somewhere other than the meager amount
> currently allocated to AGI...
>

Again, I agree with you. As I explained, AGI is but one of several
disciplines that must all work together to achieve success for any of them.
AGI is certainly not overfunded, so robbing from it wouldn't be good.
However, it just can't succeed on its own.

>
> Of course knowing a lot more about the brain would be useful for AGI, but
>
> A) I have no reason to believe that AGI is unachievable in the absence of
> this info
>

You also have no reason to believe that AGI is achievable in the absence of
this info. There are many opinions about this, and yours is in the
minority. You could easily blow away your detractors to encouraging the
development of brain diagramming systems to "guarantee" your eventual
success. Without this, it will always by your minority opinion against
everyone else's opinions. You just can see this.

Suppose for a moment that you had good reason to KNOW for a moral certainty
that brain diagramming couldn't possibly work. It would STILL be in your
interest to promote it!!! Why? Because it would provide an apparent
guarantee of eventual AGI success, which would give your prospective
investors a warm fuzzy feeling. Of course, this presumes that the money
that funds it comes out of someone ELSE'S pocket.

>
> B) I don't see any clear argument that drastically better brain imaging is
> going to be available rapidly, no matter what we do
>

Again, my diagramming proposal seems to provide that argument.

>
> C) I personally am way better trained for AGI work than brain imaging
> work, so for me to focus on the latter instead I'd need to have a really
> really really strong reason to think it would pay off...
>

Again, I agree.

I might understand if you reject what I am saying, so long as you show some
understanding how brain diagramming provides at least a perceived guarantee
of eventual success for AGI, something that AGI is now sadly lacking. In
the present environment it is tough to find funding for AGI because of this
present lack of any guarantee of eventual success - but you would know
about this better than I.

Here, you appear to be turning down a free insurance policy, and your
prospective investors are investing in other things that DO appear to have
a guarantee of eventual success. You express reasonable concerns whether
the insurance is perfect, but insurance is never perfect, and doesn't need
to be perfect. Perhaps you could explain why this isn't as dumb as it
sounds?

Steve



-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com

Reply via email to