[ExI] Shared "Mind" Database & AI Development

[Dennis May]

I wrote:

> One of the largest challenges in the development
> of AI is financial. Some ten years ago I
 estimated
> my favored approach would take some 2 billion
> dollars in lab equipment and infrastructure plus
> the budget to keep a minimum of 50 people
> involved - call it 100 million per year for 10-15
> years. Adjust for current pricing.
 
Alan Grimes wrote:
 
> That estimate
 seems rather pessimistic to me. What 
> is your reasoning?
http://en.wikipedia.org/wiki/Monolithic_microwave_integrated_circuit

My experience in the costing is primarily related to having 
worked on the MMIC development program as well as
graduate classes in sensors, physics based information theory,
and assorted optics.  The laboratory where I worked could
start from GaAs raw stock and process through to finished
MMIC devices - as well as design devices and circuits.
 
http://en.wikipedia.org/wiki/Active_Electronically_Scanned_Array
 
MMIC devices are used to form phased array radars also developed
at the lab where I worked.  The analogy being MMIC's act as the
neurons of the phased array radar.
 
Other analogies are phased array solid state lasers and massively 
parallel computers composed of many individual serial processors
- also worked on where I went to school.
 
The AI design I have in mind would require a lab with similar
capabilities plus the ability to fabricate optical devices.
 
I would consider the pricing I did to be a bargain.  

Alan Grimes wrote:

> What do you think your odds of success are?

It is my opinion that given the funding and good
management a near certainty it could be done in the
time frame outlined as long as government doesn't
interfere.  It could even be accomplished within a
government program if they returned to a management
R&D structure more like the 1950s to early 1960's 
when choking to death on paperwork [excessive 
overhead, regulation, and external and internal 
micromanagement] was not as much  an issue.

Alan Grimes wrote:

> What is your basic approach?

The key is the neuron equivalent.  In my approach
the processor communicates optically with other
processors.  The processors are in an optically
friendly 3-D support structure where they can
communicate directly with thousands of other
processors.

I wrote:

> This concept
 would be a continuation
 of the idea
> of a series of minor human upgrades - possibly
> involving a kind of smart phone technology and
> minor implants and added sensor
 capability.

Alan Grimes wrote:

> Why do you think now is the right time to attempt that?

We have a whole generation of people hungry for knowledge
and prolific with smart phones.  We also have a generation
of soldiers who have just been introduced to mobile sensor
displays, networking in the field, and computer on person
performing tasks for them as they work.  Similar work done
in cockpit diplays has now moved to mobile individual users.

Alan Grimes wrote:

> I would very much like to hear a more detailed proposal.

I can think of many applications.  Picking the most marketable
to detail is the trick.  

I wrote:
 
> The growth of the system
> can help pay for the AI work. The AI is
> integrated into the network as portions come on
> line. It would be expected that over the time
> frame of development there would be
> improvements in product delivery, improved
> sensors, improved implants, and improvements
> in the AI structure and ways to integrate it into
> the database.

Alan Grimes wrote:

> Do you think there
 is a market? How much
 work 
> do you think you will need to service that market?

You need to create a product in order to generate a
market.  The market for automobiles was always there
- but the equipment and infrastructure were not for 
millenia.  The devices and increasing infrastructure
will have to grow together.  You cannot plan up front
what all will be needed in great detail as it will evolve
depending on the mechanisms produced and how
the market of customers react to it.

Alan Grimes wrote:

> My concern about the type of bionic technology 
> available today is far too primitive and that if you 
> use it, you'd be boxing yourself into a dead-end 
> upgrade and would have great difficulty transitioning 
> to a more powerful technology. The body is resilient 
> but it is easily injured. A poorly designed upgrade 
> could do serious damage to both the patient and to 
> the reputation of transhumanism as a whole.

I agree which is why I support the first generation
devices using existing sensory input with subtle
human interaction - like a smart heads up display
with sound systems.  The military has already
been working on eye movement controlled heads
up displays and silent digital communication.
Integrate into that sophisticated external sensors
and wireless capabilities and you have something
well beyond smart phones but still no direct
medical issues.  The next obvious step from
there is first generation commerical implants
taking the most successful processes a step
further.

I wrote: 

> Like anything it would necessarily start small
> with a tiny number of applications, sensors,
> and implant options.

Alan Grimes wrote:

> Again, I see AI as a necessary pre-requisite 
> to a highly capable neural interface. Existing 
> neural interfaces are problematic for a number 
> of reasons. Also, they don't really interface with 
> the brain at an interesting level yet.

We can't start from the final answer.  It would be
like planning for the interstate highway system
based on experience with wagon technology prior 
to the railroads.

My view:

Step one - non-medical upgrades with external
senses and processing - more like personal
computing assistants and heads up displays.

Step two - money from step one funds implant
and AI research.  Implant trials before attempting
any marketing of product.  AI rearch continues.

Step three - 1st generation implant products come
on line.  What happens after step three and AI
being developed is too far out to project.  At that
point everything will snowball very quickly.

Dennis May
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20111023/23a07bca/attachment.html>