What if Turing thought it was equally as simple as? :
zn+1 = zn2 + c

(waiting for it...)


Conceptually speaking, suppose we let zn+1 = zn2 + c = Variable X < >;

Could be the "real" AGI problem is manifesting itself as a state of 
self-delusion where too much science believe there's not enough science around 
to do the job, yet the science is not forthcoming either? Where should we begin?


  1.  START => Does anyone have a theory of how to generate an AGI machine 
platform?
  2.   If YES, THEN what experiment has been proposed to test this theory and 
is it feasible to do so? => END LOOP; EXIT.
  3.  If NO, THEN 1.;

That's my version of a "scientifically sound" fatal loop. So, flying in the 
face of bold statements being made here about not "caring" about theorums in a 
scientific context, does anyone have a theory to share?

Isn't the departure of science the probability of possibility? Always tickles 
me pink how often scientists specialize in stressing the impossible. I find it 
to be such a contradiction in terms.

So, before you guys get overly excited about how impossible things might be, 
please state if you have a researched theory for generating an AGI machine 
platform? If you haven't, then you're not being scientifically sound, are you?

Ah! You're busy developing components for AGI you say? Against which 
architectural framework (I'm being lenient) and/or design?

To my mind, having a machine-run program, which can learn a language by rote 
does not constitute AGI, neither does having a machine that can drive a car. By 
academic definition of today, that's all AI, and even falls within the scope of 
narrow AI. Likewise, each scientific domain is a scientific domain, not AGI. 
One does not imply the other.

Colin, if you're so absolutely convinced a single level of complexity is 
required, then prove it. Show us the money. Else, keep your scientific options 
open.

And for the umpteenth time, please do describe the research approach and 
methodology envisaged for the experimental phase. These should be open to 
public criticism.

I'd love to see all that in a white paper one day, or in an academic proposal.

Robert Benjamin
________________________________
From: Colin Hales <[email protected]>
Sent: Saturday, 29 June 2019 08:38
To: AGI
Subject: Re: [agi] ARGH!!!

Hi Steve glad to see you back.

On Sat, Jun 29, 2019 at 3:58 PM Steve Richfield 
<[email protected]<mailto:[email protected]>> wrote:
Matt, Colin, et al,

The REAL underlying problem is that AGI started too soon - there just isn't 
enough known to be able to "fill in the blanks" and build generally intelligent 
systems. Neuroscience needed more time and money to provide this information.

I just delivered a working plan in another post.


However, the way things have been going, neuroscience would take FOREVER to get 
there - like centuries. I have been involved for half a century, and there sure 
hasn't been a half-century of progress.

Not as bad as that. Bee-level AGI in 15-20 years. Human? 2050-2070? I now have 
real numbers to use and it's getting feasible.


The problem with neuroscience is that they don't have the right tools - because 
their VW&R Catalog doesn't list high-resolution 3-D non-destructive UV 
microscopes that allow you to actually WATCH neurons in high-resolution 
operation, etc.

We have enough signalling biophysics knowledge to get on with it and do the 
science properly for a change (see my tome-responses to matt).

I figured out how to build the microscope. Everyone who has looked at my 
proposal perceives a different problem - usually while dismissing the problems 
perceived by others. The bottom line, if you embrace the idea that at least 
half the people will be at least half right, is that the microscope will do 
INCREDIBLE things - but like all laboratory instruments it will have 
limitations that can probably be worked around, like it might not be able to 
functionally diagram entire brains, etc. Further, we can't know just what those 
limitations will be until we build it.

There have been proposals by others for new tools to open up neuroscience, like 
machines to automatically serially section small (mouse?) brains for scanning, 
etc. The challenge here is to cut thousands of sections without damaging ANY of 
them.

Colin, what sort of tool(s) can YOU imagine to facilitate digging out how 
things work?

The project plan is a new chip foundry. See the plan. You build the brain  
physics on the chip and compare nature and the chip ... good old fashioned 
science. It will discover things for wet neuroscience to look for ... for 
example ... the exact location of transporter protiens and ion channels in the 
membrane. Stuff like that.


The point here is that there seems to be NO money to build this stuff, because 
there is no present market, because there is presently NOTHING like this stuff, 
etc. In short, the ENTIRE field of venture investing is fucked - and is taking 
out AGI in the process.

It's been screwed over by computer science. We just need to get the science 
configured correctly.


All that seems to be needed are a few million dollars and a couple of years of 
time by some bright guys, like some of the folks on this forum. Google spends 
more that that on coffee.

Where is Kurzweil? Where is Bazos? Where is Gates? Ben has started a bandwagon 
going nowhere other than AGI Winter, and all the big-money guys seem to be 
riding that bandwagon.

and ... I am sitting here watching this slow motion train wreck.

Depressing.

Boy am I with you! I have been watching this strange debacle full time for 20 
years.


Does anyone here have a few million dollars to spare - to make quadrillions of 
dollars in the future?

Steve

My estimate AGI #1, a dog sized robot with a bee-sized real AGI brain is 
US$2-3Billion over 10-15 years. A hundreds of millions a year for a decade? 
Like I said a better bet than UBER.

Can we keep talking about the science? It might cheer you up a bit! :-)

Colin



On Fri, Jun 28, 2019 at 10:33 PM Basile Starynkevitch 
<[email protected]<mailto:[email protected]>> wrote:

On 6/29/19 7:06 AM, Basile Starynkevitch wrote:
>
> On 6/29/19 5:21 AM, Alan Grimes wrote:
>> [email protected]<mailto:[email protected]> wrote:
>>> An AGI megaproject would require thousands of people working during
>>> several decades on it. Costing tens of billions of US$ or ???.
>>>
>>> As far as I know, it has not started. Even worse, I cannot name any
>>> megaproject today in the information technology domain.
>>
>> uh, what planet are u on?
>>
>> http://www.oecd.org/going-digital/ai/private-equity-investment-in-artificial-intelligence.pdf
>>
>> https://www.thetechedvocate.org/six-countries-leading-the-ai-race/
>> https://www.forbes.com/sites/jeanbaptiste/2019/02/12/venture-capital-funding-for-artificial-intelligence-startups-hit-record-high-in-2018/#11ad12a141f7
>>
>> https://www.statista.com/statistics/607716/worldwide-artificial-intelligence-market-revenues/
>>
>> https://secure.wealthdaily.com/187581?device=c&keyword=ai%20intelligence&gclid=CjwKCAjw9dboBRBUEiwA7VrrzTkh6YVG_yRrATxn6eInrUs-oMooia6fWUg5_JVCuGQnrulIQwt1ZhoCY28QAvD_BwE
>>
>> https://www.google.com/search?sa=N&q=global+ai+investment&tbm=isch&source=univ&ved=2ahUKEwjombvw3Y3jAhWKneAKHcNnAkc4ChCwBHoECAUQAQ&biw=1060&bih=742
>>
>>
>
> But AI is not AGI, it today is mostly big data and machine learning.
>
>
> I mentioned AGI, not AI. That makes a huge difference.
>
>
> Notice that what was called AI in the 1980s (the goal -or dream- of
> achieving human like intelligence in some artificial system) is today
> called AGI. This is a change of terminology.
>
>
> And, as far as I know, there is no today a single software system on
> which tens of thousands computer scientist are working continuously
> for an entire career, e.g. several decades.
>
> Again, compare the current large AI projects to something like ITER,
> or the Apollo program. We have no -single- that large and that
> ambitious software project on Earth.
>

And while I did mention The Mythical Man-Month previously, I forgot to
mention an important phrase from it: If one woman can make a baby in 9
months, 9 women won't make a baby in 1 month.


The equivalent is probably true for AGI, replacing months by decades. A
hundred of ambitious AI projects won't advance AGI, even if run in
parallel. Exactly like ten thousands of rocket clubs firing rockets to
50 km don't improve any human to Mars space mission.

We don't have the social & political structures to even start an
ambitious single software project with 10000 computer scientists working
for several decades. And we might not even know what organizational
structure could make such a project possible.


So I don't expect AGI to come soon.


--
Basile STARYNKEVITCH   == http://starynkevitch.net/Basile
opinions are mine only - les opinions sont seulement miennes
Bourg La Reine, France



--
Full employment can be had with the stoke of a pen. Simply institute a six hour 
workday. That will easily create enough new jobs to bring back full employment.

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-Mddadd648b86d2de8421920c9>

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T87761d322a3126b1-M07178690366981ef1b0a4987
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to