--- Russell Wallace <[EMAIL PROTECTED]> wrote:

> *nods* Would it be accurate to say that you see the encoding of
> specific knowledge as the difficult and expensive thing, then, and so
> you're trying to create an environment that maximizes the resources
> that can be brought to bear on it, while also providing as many
> opportunities as possible for incremental progress (individual
> modules) to be rewarded and reused?

Yes, my goal is to make communication easier (human-human,
machine-machine, and human-machine).  I believe that connecting lots of
narrow AI in a useful way makes AGI.  A lot of the narrow AI that needs
to be solved to make this work has already been solved.  Deploying the
communication network is probably the biggest hurdle.  It may take a
decade.  People will need a compelling reason to adopt it, and there
won't be one until there are already lots of users.

The alternative approach is monolithic human-level AGI.  Right now I
estimate human level AGI will cost on the order of US$1 billion, split
evenly between hardware (10^6 PCs to simulate a brain sized neural
network) and software (10^6 lines of code, roughly the complexity of
the human genome).  The hardware cost will come down and the software
can be copied.  But that is still doing it the hard way.  We don't
really want artificial humans.  We want machines that do what we tell
them to.  They do need human level skills like language and vision to
do that.  But these capabilities can be developed incrementally -- good
enough for the job they do -- and put to good use before the monolithic
AGI problem is solved.

A good communication network and copying knowledge might eliminate 90%
of the training costs of distributed AGI compared to training humans. 
But the last 10% is still going to cost $hundreds of trillions, a few
years of world GDP, assuming you have the 10^10 human brain equivalents
needed to run the economy.


-- Matt Mahoney, [EMAIL PROTECTED]

-------------------------------------------
agi
Archives: http://www.listbox.com/member/archive/303/=now
RSS Feed: http://www.listbox.com/member/archive/rss/303/
Modify Your Subscription: 
http://www.listbox.com/member/?member_id=8660244&id_secret=101455710-f059c4
Powered by Listbox: http://www.listbox.com

Reply via email to