Re:
I want to do it.. the motivatin for the scientists. they are the ones that proposed tha accelerator,, lets assume a brooklyn project, with no holds barred. how much, who and for how long?
----- Original Message ----
From: Ben Goertzel ben@goertzel.org
To: J. Epstein jeeproject@yahoo.com
Sent: Tuesday, April 15, 2008 6:29:50 AM
Subject: Re:
Hmmm...
You're searching for a motivation for YOURSELF, or for the scientists involved??
To me, working on stuff like curing disease (which I've worked on via
Biomind) and flying to the moon is short-sighted. The goal is to
create an AI scientist who can solve these problems vastly better than
we can. Ultimately maybe to become this AI scientist, those of us who
want to. These things sound "out there" but it seems very clear they
are physically possible; they sound "out there" to us because of our
cultural preconceptions.
Once you've created a greater-than-human mind, all sorts of other
things become possible.
I guess you've read Ray Kurzweil's recent book? I tend to concur with
his sentiments, though I think he whitewashes the possible bad
outcomes of advanced tech, and underestimates the odds of a
Singularity coming sooner than his schedule due to a radical
breakthrough. (I have gotten to know Ray a bit in recent years, by
the way ... and also, had an article on this theme in Artificial
Intelligence journal ... I thought it was an interesting sign of
change that this prestigious mainstream journal was willing to publish
articles debating Ray's ideas about the future of AI and society.)
I'll tell you how I view the next few decades of human history, which
may be a bit different than how Ray or anyone else views them. I tend
to agree with Ray that we are likely moving toward some kind of
technological Singularity ... that certain extremely impactful kinds
of core tech are gonna keep advancing faster and faster with more and
more dramatic consequences. But I'm not nearly as convinced as he is
that the outcome will be good. Venter's and related work on
engineered micro-organisms scares me ... eventually some nutcases will
create biopathogens and annihilate all or nearly all of us. And then
there's nanotech. I think our best hope of surviving this century is
to create advanced AI's that enable abundance for all (and the curing
of human psychological problems), so that peoples' motivations to use
technology in nasty ways decreases ... and of course so that the AI's
themselves can prevent these destructive uses before they happen.
A bit science-fiction-ish, I realize ... but so would the Web have
seemed 50 years ago or even 30 ...
Talk about biomedicine ... I think Aubrey is right that eventually
we'll cure the disease of aging ... but if science is proceeding at
the pace of human thought, it may take a century of research. An AI
scientist could design the right experiments way faster and solve the
problem in a few years ... bingo, we don't have to die anymore, unless
we want to (and I don't ... )
From what I have seen (for instance at AGI-08) there are loads of
bright young scientists with a passion for building thinking machines,
but there is minimal funding for that sort of research, so people get
sidetracked onto what they can get paid to do. Folks start out in the
AI field wanting to build thinking machines but by the time they 5
years past their PhD, they're side-tracked onto what DARPA or NSF are
willing to fund. And even those who maintain the passion are
struggling. Stan Franklin (co-chair of AGI-08) is 75 years old and
his health is failing ... he entered AI after a first career in
mathematics ... he has some pretty good ideas about AI (overlapping a
lot with my own) and has tried to push his lab in this direction ...
but most of their work wound up going into building interactive
scheduling systems for the Navy, as that's what they managed to get
funding for.
Society will pay to build a huge-ass particle accelerator, without any
immediate motivation. There is no enemy nor prize telling us we need
to build a Large Hadron Collider now ... but we're willing to spend
billions on it because it's a Big Metal Box and our monkey brains are
wired to appreciate large impressive-looking objects.
In short: the increasing pool of researchers who "get it" don't need
any special motivation to want to push really hard toward creating
thinking machines at the human level and beyond ... any more than
particle physicists need some prize or other special motivation to
spend their lives trying to understand the foundations of the
universe.
Hey, if you want to help AI but can't summon enough enthusiasm to
launch the Brooklyn Project, consider donating to our open-source AI
project OpenCog.org ..... While I'm still pumped about the role my
startup firm can play, I'm increasingly convinced the open-source
approach is gonna be what gets us there to the superhuman thinking
machine. Ultimately creating a human-level AI is going to require
fewer work-hours than creating Linux has taken. My company and I can
do the hardest parts, but hundreds around the world can potentially be
leveraged to help with all the other parts, contributing their own
efforts as well as research ideas. [Profit for my company along the
way then comes, not from owning the Ai code, but from understanding
better how to leverage it for vertical apps like virtual pets... but
to me, making a profitable AI company is just a subgoal .. and far
less interesting than, er, obsoleting the paltry human mind...]
I talked to Marvin Minsky about this stuff last year when we both
spoke at the TransVision 2007 conference. He seems to have a fairly
similar overall vision to me, but I just don't feel like he has a
coherent overall design for a thinking machine (whereas I do). Push
Singh was trying to morph Minsky's ideas in that direction but, well,
his human biopsychology got in the way ;-p .... Fortunately, I do get
frustrated but not that frustrated...
Anyway, I guess you know my attitude pretty well. Maybe I'll give you
a call sometime, or if you want to meet F2F let me know.
I've made a lot of connections in the last few years such that if you
wanted to pull together an AI Brooklyn project now I could do it
(whereas in 2001 I was more-so living in a vacuum ... lately I've
taken on a lot more of a role in the academic AI community and the
futurist community as well...)
Think about it ;-)
Ben
On Tue, Apr 15, 2008 at 6:57 AM, J. Epstein jeeproject@yahoo.com wrote:
I don't like prizes ,Howeverm I'm searching for a motivation. the manhattan project had it. teams sports have it. research for a family disease, but pure Ai, seems Like __HEY LIke OK, Dude , We'll get around to it.
----- Original Message ----
From: Ben Goertzel ben@goertzel.org
To: J. Epstein jeeproject@yahoo.com
Sent: Monday, April 14, 2008 10:13:17 PM
Subject: Re:The Brooklyn Project is long overdue -- let's do it ;-)
My current approach is not exactly web-based, we're using the AI to
control virtually embodied agents in online virtual worlds.I think that {Edelman,Minsky, Abelson, Lloyd, Goldberg} are developing
fascinating ideas but are not really pushing hard toward the grand
goal, at this point...About the "AI X Prize" idea, I have actually thought about that a lot,
but it seems a hard problem. Hard to find a test that can't be
"gamed" by narrow-AI trickery... and isn't so hard (like the
Turing test) as to be useless for motivating immediate-term
development...I like this Recognizing Textual Entailment challenge
http://www.pascal-network.org/Challenges/RTE/
and I thought of extending it into a Recognizing Situational
Entailment challenge, where you show an AI a situation in a virtual
world, and let it interact within that situation to gather information
... and then you ask it binary questions about the world, and see if
it gets the answers right.This would be a very hard test and a wizzy one too ... and I don't
think it could be gamed.But ultimately, I feel prizes like that work best when the goal is (as
with the space-flight X prize) to push toward a lower-cost
implementation of well-understood ideas. For cases like building a
thinking machine where the foundational principles still need to be
verified, I'm not sure a prize is the most apropos approach. I like
the Brooklyn Project better ;-)ben
On Mon, Apr 14, 2008 at 11:03 PM, J. Epstein jeeproject@yahoo.com wrote:
my office # is 212 750 9895'' . I'm thinking of the Brooklyn Project again, i've funded some AI nothing that interesting . ,Edelman,Minsky, Abelson, Lloyd, Goldberg. I'm considering offering an X like prize for a thinking learning/ machine, I can't yet decide on the rules for winning. Not a Turing test,- I'd like to have your suggestions., is your archictecture web based are have you grown more ambitious. I followed some of your AGI.. glad to hear you found more than programmers in Brazil, -
----- Original Message ----
From: Ben Goertzel ben@goertzel.org
To: J. Epstein jeeproject@yahoo.com
Sent: Monday, April 14, 2008 8:40:57 PM
Subject: Re:Hello Jeffrey,
Good to hear from you!
I'm living in Maryland near Washington DC (moved here from New Mexico
in 2002).First let me address a bit of ancient history: I felt bad that we did
not correspond much during 2001-2002 when you funded my work via U. of
New Mexico. What happened was a combination of odd flukes... 1)
somehow your assistant got the idea that I was at the Santa Fe
Institute rather than UNM, and kept trying to contact me there; 2)
shortly before I left UNM in 2002 I found some messages from your
assistant on a voicemail at UNM that I had never set up or listened to
before. Anyway the best way to contact me is always email.... (I did
try to email you various work summaries during that period but a
dialogue never got going.)Anyway, as for the present...
In the time since we last talked, I've
Put huge effort into refining my AI design (an effort that largely was
completed about 2 years ago, though I'm still working on improving
various aspects).The rough-draft manuscript I showed you way back has evolved into
a) a book on philosophy of mind called The Hidden Pattern, published
2006 by Brown Walkerb) a book on probabilistic logic, called Probabilistic Logic Networks,
to appear in a couple months, published by Springer Verlagc) a not-yet-published book-length doc giving the crux of the AI
design, a slightly out of date draft of which can be found atusername: novamind
password: novamindThis is going to be edited further to form foundational documentation
for the open-source OpenCog system (see below).
Started an AI software consulting company, Novamente LLC, which now
has 20 staff and is moderately but not hugely successful,
http://novamente.netWe are now trying to transition Novamente away from being a consulting
firm and toward a focus on intelligent virtual pets, see
http://www.novamente.net/zeptopetz/ ... we just hired our first
full-time marketing person with this focus in mind...Some of the philosophy underlying the "virtual pets" as a first step
toward powerful artificial general intelligence are outlined in this
blog post:http://www.novamente.net/blog/?p=9
Begun a major initiative to make really ambitious AI work more
acceptable within the academic, corporate and gov't community.Mainly, I co-organized a conference that was at U. of Memphis in March, see
It went fabulously well, and AGI-09 will be in the DC metro area next
spring (details to be announced).We even had some frank discussion of your old notion of an "AI
Manhattan Project" in one of the conference sessions, which was cool
as there were some fairly conventional AI academics in the room,
psyched about the idea.
Quite recently, launched an open-source AI software project called
OpenCog. We got a $100K grant from the Singularity Institute for AI
(http://singinst.org) to clean up some of the core Novamente code for
open-sourcing.OpenCog was accepted as a project by Google Summer of Code, which
means Google will be funding about 11 students to spend the summer
working on the project. That will be nice and will likely be repeated
in future summers.
Oh, and as an aside, in 2002 I started a bioinformatics startup,
Biomind LLC, which has survived on government contracts and has done
some pretty cool stuff applying novel but not that ambitious AI tech
to analyzing data regarding Parkinson's, Alzhemers and other diseases
... and studying the changes calorie restriction diets induce in mice.Our AI was instrumental in making a discovery that seems likely to
lead to the creation of the first blood test for Parkinson's and
Alzhemier's. But we may need a lawsuit to get any revenue for this
work. But that's a whole other story...Anyway ... all in all ...
-- I have a way better design for a thinking machine than I did in
2001, based on the same philosophy, but simpler and smaller and more
mathematical. ;-)-- I have a team of really smart AI guys working for my consulting
company (plus two working for the OpenCog project) who understand most
of my vision for AI ... most of these guys are in Brazil but some are
scattered all over the world...-- sadly, I only have a couple of these guys working on stuff directly
related to building a thinking machine ... mostly we're occupied with
doing shorter-term, narrower AI project that bring in small amounts of
revenue and keep us fedI reckon that at the current pace we are 20-25 years from creating a
human-level AI; and that this could be accelerated to about 5-9 years
with a "modest" level of funding, of say $1M/year devoted to nothing
but thinking-machine-building.But the world remains rather uninterested and over-skeptical about the
possibility of achieving human-level AI in the near-term, which
remains incredibly frustrating to me, but I'm slowly learning to live
with it ... and anyway, the work does proceed, albeit way too slowly.And in the meantime, I've turned 41, moved on to my second wife (a
brilliant, mid-20's Brazilian girl) ... and my oldest son is a
sophomore in college studying math, CS and physics!! Human life has a
way of proceeding forwards ;-pAs an aside, I wonder if you've kept abreast of my friend Aubrey de
Grey's work on life extension ... http://sens.org .... I think you
would get a kick out of it and be able to appreciate its value. There
are of course close relations between Aubrey's vision and our work
applying AI to bioinformatics.It would be great to get together and talk about AI again face to
face. Just let me know when/if you're interested. New York is a
quick train ride from DC and I do get up there fairly often anyways
for various reasons.In fact I have had two recent meetings in NY with a banker named James
Benedict, who is trying to line up investment money for Novamente LLC
from wealthy individuals he knows whom he thinks may be motivated by
the combination of our longer-term thinking-machine plans, and our
short-term virtual-pets business avenue.Let me know...
My cellphone is 240-505-6518 ... but I check that voicemail only
erratically, and email remains by far the best way to contact me...Y'know ... if you'd funded the AI Manhattan project back in 2001, we'd
be a lot closer to having a human-level thinking machine than we
are now. I've grown up a lot as a human being since we last talked,
but, my confidence about the pathway to AI has just grown greater as
I've thought through the issues, and refined my ideas, more and more.Anyways...
I hope you're doing well also! ... and let me know if you'd like to
get together and chat...thx
BenOn Mon, Apr 14, 2008 at 7:24 AM, J. Epstein jeeproject@yahoo.com wrote:
where are you and what are you up to --------------Jeffrey Epstein
--
Ben Goertzel, PhD
CEO, Novamente LLC and Biomind LLC
Director of Research, SIAI
ben@goertzel.org"If men cease to believe that they will one day become gods then they
will surely become worms."
-- Henry Miller--
Ben Goertzel, PhD
CEO, Novamente LLC and Biomind LLC
Director of Research, SIAI
ben@goertzel.org"If men cease to believe that they will one day become gods then they
will surely become worms."
-- Henry Miller
--
Ben Goertzel, PhD
CEO, Novamente LLC and Biomind LLC
Director of Research, SIAI
ben@goertzel.org
"If men cease to believe that they will one day become gods then they
will surely become worms."
-- Henry Miller