RE: Singularity Institute - update

From: Ben Goertzel (ben@goertzel.org)
Date: Sat May 03 2003 - 14:12:12 MDT


> I also believe that open sourcing some of the components that
> are generally useful will be a much faster way of getting those
> components built than waiting for everything including funds to
> come together to do the job on a closed source basis. A few
> good architects to create component subprojects and manage them
> as OSS would be easier to come by than many more master
> programmers dedicating their lives to a more closed effort with
> little or no working capital or compensation until a hopefully
> happy Singularity.
>
> I think there is a false assumption that useful people resources
> need to understand everything about FAI in near Eliezer-like
> depth before they can do a lot of very useful and needed work.
>
> Given that we believe both that getting to FAI is utterly
> crucial and every day until that point is full of huge risks to
> all humanity and that FAI is difficult on technical and
> theoretical levels, it seems to me we need to divide and conquer
> the problem as much as possible. Otherwise stasis will prevail.

It is true, there are useful things that could be done on an OSS basis to
help us along the path to AGI

For example, Novamente could use

*
a robust visualization tool for visualizing and navigating complex networks
[including labeled weighted hypergraphs, both small and huge ones].
Existing tools like Tulip are a start, but no existing tool really fits the
bill in terms of visualizing the insides of a Novamente instance.... And
the same tool that could visualize the inside of Novamente could visualize
the inside of a lot of other would-be AI systems, e.g. Peter Voss's A2I2
system which is a network of a different sort (more lik ea neural network
than Novamente)

*
an easy to use "distributed processing framework" allowing Novamente or
other AGI systems to farm out certain types of processing across millions of
processors, distributed.net / SETI@home style.... Again, such tools exist,
but when you look at the details none of them are flexible enough to use
directly off the shelf for Novamente AGI purposes, or for other AGI purposes
either I suspect

*
software that takes currently available perceptual processing software (e.g.
sound and vision processing software) and outputs the results in a
Novamente-friendly framework (say, the right sort of XML), and causes it to
accept control commands in a simple way.... [Eventually Novamente will do
its own perceptual processing, but for initial experimentation in this
domain, hooking it up to existing perceptual software may be an interesting
"scaffolding" step along the path...]

None of these is a small project. All of these would likely benefit
multiple AGI projects, not just Novamente. All could be developed on an OSS
basis by a distributed team of enthusiastic part-timers.

So, if anyone feels like working on one of these things, let me know and we
can collaborate ;) The catch is that I do not have time to actively
organize and manage an OSS effort in any of these areas, only to help out
with general direction-giving and intermittent technical problem-solving.
So what is needed is not merely some enthusiastic programmers, but a
programmer/team-leader, which is a different thing.

These sorts of projects require significantly less AI and math knowledge
than working on Novamente itself .. the learning curve is definitely less
(and the learning curve for doing significant work on Novamente is a damn
tough one...)

-- Ben Goertzel



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:42 MDT