From: Ben Goertzel (ben@goertzel.org)
Date: Fri Nov 08 2002 - 08:39:00 MST
Gary Miller wrote:
> I would be less interested in seeing a blog than seeing a well
> maintained project plan in enough detail to show me that the research
> plan is in fact concrete enough that it can be broken into definable
> steps with periodic milestones.
The Novamente AGI project (www.realai.net) has a concrete research plan,
broken into definable steps with periodic milestones.
We are not maintaining formal project-planning documentation for our AGI
project at this point, though we are maintaining such documentation for our
Novamente-based narrow-AI commercial software projects.
It would be quite easy to put such documentation together, however, if there
were a need (such as a likely funding source).
> This has the added benefit of allowing volunteers to step up to specific
> tasks and perhaps donating skill and expertise as well as money.
We are using volunteer software engineers (one full-time, a few other
part-timers); and we also have a full-time engineering team who are largely
devoted to Novamente-based commercial projects, but are along the way doing
some coding pertinent to the AGI effort.
This is a lot better than making no progress at all, yet -- I can't deny
it -- it's a lot worse than having funding to pay a comparable team to work
on AGI full-time.
The compensating factor is that the Novamente-based commercial projects are
also really interesting, though not as important in a grand sense as AGI.
(Our largest project, Biomind, involves applying Novamente to analyze
genetic and proteomic data.)
> I realize that research efforts often cannot be specific as to when or
> if a given long-term objective will be accomplished, but the project
> plan could list immediate 1-3 month goals, medium range goals 4-12 month
> goals and long term.
I, and many other AGI researchers, understand how to plan a project. The
reason funding for AGI is hard to achieve is not that researchers in the
area lack software designs or organizational skills; the reason is that AGI
is viewed as an unrealistic, pie-in-the-sky goal by the scientific "powers
that be."
> I am fairly new to the list and have read some postings regarding ethics
> of the AI, some regarding the funding, but thus far nothing has really
> demonstrated to me how this large project is to be attacked.
I have started an e-mail list called the "AGI List" that is more narrowly
focused on artificial general intelligence and means of achieving it. Go to
www.realai.net and scroll to the bottom of a page and there's a place to
sign up, if you wish.
> Let me bring up another funding idea though. Think of some aspect of
> the early project phases that could be made interesting to National
> Science Foundation. It sounds like some of the principles here may have
> the scientific credentials to pull off an NSF grant. Doing a grant
> proposal does require enough of a plan to instill credibility but then
> so does fundraising,
I know of two other leading AGI researchers, besides myself, who had NSF
proposals rejected this year.
The rejection letter for my proposal specifically stated that they were more
interested in funding proposals aimed at specific application areas, and did
not want to fund general intelligence research.
My scientific credentials are not so bad, and one of the others whom I'm
referring to has very good scientific credentials also (he has more
mainstream AI publications than I do, but fewer publications overall)...
Next year I'll submit an NSF application for bioinformatics work instead,
because I think that will have a prayer of success.
We're doing bioinformatics with our Novamente codebase, so if this work gets
NSF funding, it will indirectly benefit our AGI project.
You have to understand the history of AI. Too many AI researchers falsely
claimed they were just around the corner from creating a general
intelligence. This happened so many times that now anything vaguely hinting
at general intelligence is greeted with scathing skepticism by the majority
of the AI research community and by the gov't research funding
establishment.
Once there is a big, publicized proto-AGI success, the NSF will be all over
AGI with funding. Not before. That's my prediction anyway...
-- Ben G
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:41 MDT