On Thu, Aug 15, 2013 at 5:46 PM, Steve Richfield
<[email protected]> wrote:
> The BIG question is: Who get the quadrillion dollars?

It cost $1 quadrillion in constant 2013 dollars to advance our global
economy to its current state from where it was 20 years ago. Where did
that money come from?

My AGI design includes a funding model. When you have a lot of narrow
experts and a network that routes messages to the right ones, you have
AGI, as in automating what you would otherwise have to pay people to
do. Individuals have a vested interest in selling these services,
either doing work or knowing who can do it. In an economy where
information has negative value, you have to invest in knowledge and
computing power to get people's attention. This is nothing new. It is
how the internet already works. Already, tens of trillions of dollars
have been spent to build it and make it smarter, in spite of the fact
that nobody has that much money.

> Either a social reformation will occur before AGI appears, or most of the
> world's population, probably including you and whoever else invents AGI,
> will starve as a result. As things now stand, I see AGI as MUCH more of a
> threat than a promise.

I guess you haven't noticed that machines already do most of the work.
Rather than people starving, agriculture has gone from almost all of
the world's economy in 1800 to 6% today (and 1.5% in the U.S.). Now,
food is so cheap that in developed countries, the poor are more likely
to be overweight than the rich.

> Further, if you read my prior postings where I explained to Ben the VERY
> negative value of near-term AGI predictions on everyone working in the
> field, you would see the negative value of your own posting. Hyping AGI is
> DANGEROUS to us all, including you.

I have been lurking on the OpenCog mailing list for a few years. There
is no chance that it will ever be powerful enough to be considered
dangerous. You don't build AGI by trying to solve all of the problem
at once. If you want to get funded, your system has to do work. We
hire people to do work because currently only humans can solve many
hard problems in language, vision, robotics, art, and predicting human
behavior. This is different than the problem that OpenCog is trying to
solve: building artificial human minds with all of their weaknesses
like emotions and cognitive limitations. They are doing no new
research into these hard problems, and are a long way from solving
them. MOSES (an evolutionary learner), DeSTIN (neural vision), ReLex
(rule based parser), NatGen (text output), and AtomSpace (hypergraph
based knowledge representation) are toys that lack the vast computing
power and knowledge bases needed to solve these hard problems. Most of
the work over the last several years has been trying to integrate
these components and maintaining a huge code base in several languages
that is a major effort to install and then doesn't do anything once it
is running.

But I do have to commend Ben for the effort. Nobody else on this list
is even doing any work at all in AGI.

-- 
-- Matt Mahoney, [email protected]


-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to