From: Christopher McKinstry (cmckinst@eso.org)
Date: Tue Jul 03 2001 - 18:07:21 MDT
> >On Sun, 1 Jul 2001, Chris & Jessie McKinstry wrote:
> >
> >> 1 - GAC is a black box. I have made no disclosures on what it uses for
> >> pattern matching (not that it should matter if you place any value in
> >> the Turing Test which bars knowledge of the internals of a system.)
> >> But, for the record, I am doing experiments now with SOMs and SRNs.
> >
> >Please expand the acronyms. Google is remarkably useless when confronted
> >with either of them or a combination of them. To wit:
> >http://www.google.de/search?num=100&hl=en&safe=off&q=SRN+SOM
Sorry. SOM= self-organizing mao (Kohonen), SRN= Simple recurrent network
(Elman)
> >> 2 - The primary purpose of GAC is to build a fitness test for
> >> humanness in a binary response domain. This will in the future allow
> >
> >Boolean metrics would seem rather contrived. We're anything but boolen.
> >
> >> GAC to babysit a truly evolving artificial consciousness, rewarding
> >> and punishing it as needed.
> >
> >Hard edges don't give you a useful fitness gradient. I'm curious as to
> >this particular decisions.
People are asked to respond in binary, but what GAC stores is actually a
continuous variable from -1 to 1 that is based on the probability of a
human responding true to that item.
> >
> >> 3 - The key to evolving anything is the fitness test. If I want to
> >> evolve a picture of the Mona Lisa, then I need a fitness test for the
> >> Mona Lisa. A good fitness test for the Mona Lisa would be a copy of an
> >> image of the Mona Lisa. To rate the quality of an evolving picture, I
> >> would just need to compare pixels. The next best thing to using an
> >> image would be a random sample of pixels; the larger the sample, the
> >> better will be the evolved copy. Right now, GAC is a 350,000+ term
> >> fitness test for humanness. At each one of those points GAC knows what
> >> it should expect it were testing an average human, because for each
> >> one of those points GAC has made at least 20 measurements of real
> >> people.
> >
> >Random online geek is hardly comparable to a CyClist.
No, but they are just as human and MUCH less expensive. Between Open
Mind and Mindpixel we've essentially duplicated the core effort of CYC's
15 year effort in less than 1 year and for millions and millions of
dollars less cost.
You can have a team of nearly zero smart people for a very large amount
of money, or a very large team of average people for nearly zero money.
The big team will always produce more data, and if you do you statistics
right, it will be of higher quality. As well, the data is framework
independent - you can process it however you choose - you can even
revalidate with expensive experts if you chose. Whatever.
The big advantage of building a big dataset is that with enough data,
you can use automatic programming to generate and test millions and
millions of potential frameworks. This is much more effective use of
resources than paying a bunch of specialists to handcode a single
framework.
> >
> >> 4 - Any contradictions in GAC are real contradictions in us. It can't
> >> believe anything that hasn't been confirmed by at least 20 people.
> >>
> >> 5 - GAC is science. Over 8 million actual measurements of human
> >> consensus have been made. There are at least two other projects that
> >
> >Yes, I contributed about 100 of them. I was trying, but both the absence
> >of domain allocation and leads and the procrustian nature of the
> >submission data made 99% of that pure crap. I suspect the other
> >contributors fared little better.
> >
> >> claim to be collecting human consensus information - CYC and Open Mind
> >> - neither has actually done the science to verify that what is in
> >> their databases is actually consensus human fact. It's all hearsay
> >> until the each item is presented to at least 20 people (central limit
> >> theorem.)
> >
> >Dunno, both your methods are obviously flawed, and your attitude does seem
> >to have a few problems. Mindpixel is miles away even from Cyc, which at
> >least was done by a number of smart, experienced experts, using a
> >comparatively rich framework.
You can use any framework you like. But the core ontologies are binary
rules in all cases, including CYC. But CYC's rules aren't science - they
are conjecture, which I think is a problem.
A way of thinking about the whole Mindpixel approach is to think of
signal processing: Think of complex thought as being like a complex wave
form and binary propositions being like component sine waves - you even
get to use the same math, which is very, very cool. And quite telling as
well.
Chris
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:08:28 MST