> From: Ben Goertzel [mailto:[EMAIL PROTECTED]
>
> Somewhat similarly, I've done coding on Windows before, but I dislike
> the operating system quite a lot, so in general I try to avoid any
> projects where I have to use it.
>
> However, if I found some AGI project that I thought were more promis
Mark,
"...and that the (actually explicit) assumption underlying the whole
scientific method is that the same causes produces the same results.
Comments?"
It seems like a somewhat weaker assumption *could* work; namely, "the
same causes produce the same probability distribution on effects".
This
>> -- truly general AI, even assuming the universe is computable, is impossible
>> for any finite system
Excellent. Unfortunately, I personally missed (or have forgotten) how AIXI
shows or proves this (as opposed to invoking some other form of incompleteness)
unless it is merely because of the
I am arguing by induction, not deduction:
If the universe is computable, then Occam's Razor holds.
Occam's Razor holds.
Therefore the universe is computable.
Of course, I have proved no such thing.
Yep. That's a better summation of what I was trying to say . . . .
Except that I'd like to bri
>> People seem to debate programming languages and OS's endlessly, and this
>> list is no exception.
Yes. And like all other debates there are good points and bad points.:-)
>> To make progress on AGI, you just gotta make *some* reasonable choice and
>> start building
Strongly agree. Ot
Ah. An excellent distinction . . . .Thank you. Very helpful.
Would it then be accurate to saySCIENCE = LEARNING + TRANSMISSION?
Or, how about,SCIENCE = GROUP LEARNING?
- Original Message -
From: "Russell Wallace" <[EMAIL PROTECTED]>
To:
Sent: Saturday, October 2
AIXI shows a couple interesting things...
-- truly general AI, even assuming the universe is computable, is impossible
for any finite system
-- given any finite level L of general intelligence that one desires, there
are some finite R, M so that you can create a computer with less than R
processi
--- On Sat, 10/25/08, Mark Waser <[EMAIL PROTECTED]> wrote:
> Ummm. It seems like you were/are saying then that because
> AIXI makes an
> assumption limiting it's own applicability/proof (that
> it requires that the
> environment be computable) and because AIXI can make some
> valid conclusions
MW, mine was an editorial reply to what struck me as a superficial
pronouncement on a subject not amenable to treatment so cursory. But I
like it less now, and I apologize.
Eric B
---
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed:
>
> Strong agreement with what you say but then effective rejection as a valid
> point because language issues frequently are a total barrier to entry for
> people who might have been able to do the algorithms and structures and
> cognitive architecture.
>
> I'll even go so far as to use myself as
Agree, that was not a useful response ...
On Sat, Oct 25, 2008 at 5:50 PM, Mark Waser <[EMAIL PROTECTED]> wrote:
> Surely a coherent reply to this assertion would involve the phrases
>> "superstitious", "ignorant" and "FUD"
>>
>
> So why don't you try to generate one to prove your guess?
>
> Are
On Sat, Oct 25, 2008 at 11:14 PM, Mark Waser <[EMAIL PROTECTED]> wrote:
> Anyone else want to take up the issue of whether there is a distinction
> between competent scientific research and competent learning (whether or not
> both are being done by a machine) and, if so, what that distinction is?
OK. A good explanation and I stand corrected and more educated. Thank you.
- Original Message -
From: "Abram Demski" <[EMAIL PROTECTED]>
To:
Sent: Saturday, October 25, 2008 6:06 PM
Subject: Re: [agi] constructivist issues
Mark,
Yes.
I wouldn't normally be so picky, but Godel's
So where is the difference
There is no difference.
Cool. That's one vote.
Anyone else want to take up the issue of whether there is a distinction
between competent scientific research and competent learning (whether or not
both are being done by a machine) and, if so, what that distinction
Mark,
Yes.
I wouldn't normally be so picky, but Godel's theorem *really* gets misused.
Using Godel's theorem to say made it sound (to me) as if you have a
very fundamental confusion. You were using a theorem about the
incompleteness of proof to talk about the incompleteness of truth, so
it sound
William,
On 10/24/08, William Pearson <[EMAIL PROTECTED]> wrote:
>
> I can't see a way to retrofit current systems to allow them to try out
> a new kernel and revert to the previous one if the new one is worse
> and malicious, without a human having to be involved.
Digging into my grab bag of lo
--- On Sat, 10/25/08, Mark Waser <[EMAIL PROTECTED]> wrote:
> > Scientists choose experiments to maximize information
> > gain. There is no
> > reason that machine learning algorithms couldn't
> > do this, but often they don't.
>
> Heh. I would say that scientists attempt to do this and
> machi
Ummm. It seems like you were/are saying then that because AIXI makes an
assumption limiting it's own applicability/proof (that it requires that the
environment be computable) and because AIXI can make some valid conclusions,
that that "suggests" that AIXI's limiting assumptions are true of the
--- On Sat, 10/25/08, Mark Waser <[EMAIL PROTECTED]> wrote:
> > The fact that Occam's Razor works in the real world
> > suggests that the
> > physics of the universe is computable. Otherwise AIXI
> > would not apply.
>
> Hmmm. I don't get this. Occam's razor simply says
> go with the simplest
Surely a coherent reply to this assertion would involve the phrases
"superstitious", "ignorant" and "FUD"
So why don't you try to generate one to prove your guess?
Are you claiming that I'm superstitious and ignorant? That I'm fearful and
uncertain or trying to generate fearfulness and uncert
Which "faulty" reasoning step are you talking about?
You said that there is an alternative to ad hoc in optimal approximation.
My request is that you show that the optimal approximation isn't going to
just be determined in an ad hoc fashion.
Your absurd strawman example of *using* a bad solut
> I'll even go so far as to use myself as an example. I can easily do C++
> (since I've done so in the past) but all the baggage around it make me
> consider it not worth my while. I certainly won't hesitate to use what is
> learned on that architecture but I'll be totally shocked if you aren't
>
On Sun, Oct 26, 2008 at 1:19 AM, Mark Waser <[EMAIL PROTECTED]> wrote:
>
> You are now apparently declining to provide an algorithmic solution without
> arguing that not doing so is a disproof of your statement.
> Or, in other words, you are declining to prove that Matt is incorrect in
> saying tha
Vladimir said> > I pointed out only that it doesn't follow from AIXI that
ad-hoc is justified.
Matt used a chain of logic that went as follows:
AIXI says that a perfect solution is not computable. However, a very
general principle of both scientific research and machine learning is
to favor si
Scientists choose experiments to maximize information gain. There is no
reason that machine learning algorithms couldn't do this, but often they
don't.
Heh. I would say that scientists attempt to do this and machine learning
algorithms should do it.
So where is the difference other than in
The fact that Occam's Razor works in the real world suggests that the
physics of the universe is computable. Otherwise AIXI would not apply.
Hmmm. I don't get this. Occam's razor simply says go with the simplest
explanation until forced to expand it and then only expand it as necessary.
How
>> Anyway language issues are just not the main problem in creating AGI.
>> Getting the algorithms and structures and cognitive architecture right are
>> dramatically more important.
Strong agreement with what you say but then effective rejection as a valid
point because language issues freque
On Sun, Oct 26, 2008 at 12:17 AM, Mark Waser <[EMAIL PROTECTED]> wrote:
>> No, it doesn't justify ad-hoc, even when perfect solution is
>> impossible, you could still have an optimal approximation under given
>> limitations.
>
> So what is an optimal approximation under uncertainty? How do you kno
Dr. Matthias Heger wrote:
...
I think humans represent chess by a huge number of **visual**
patterns. The chessboard is 8x8 squares. Probably, a human considers
all 2x2, 3x3 4x4 and even more subsets of the chessboard at once
beside the possible moves. We see if a pawn is alone or if a knight
So you're saying that if I switch to using Tarski's theory (which I believe
is fundamentally just a very slightly different aspect of the same critical
concept -- but unfortunately much less well-known and therefore less
powerful as an explanation) that you'll agree with me?
That seems akin to
No, it doesn't justify ad-hoc, even when perfect solution is
impossible, you could still have an optimal approximation under given
limitations.
So what is an optimal approximation under uncertainty? How do you know when
you've gotten there?
If you don't believe in ad-hoc then you must have a
Dr. Matthias Heger wrote:
The goal of chess is well defined: Avoid being checkmate and try to
checkmate your opponent.
What checkmate means can be specified formally.
Humans mainly learn chess from playing chess. Obviously their knowledge
about other domains are not sufficient for most beginner
Eric,
Nobody here is actually arguing that the brain is non-computational,
though. (The quote you refer to was a misunderstanding).
I was arguing that we have an understanding of noncomputational
entities, and Ben was arguing (approximately) that any actual behavior
could be explained equally wel
--- On Sat, 10/25/08, Mark Waser <[EMAIL PROTECTED]> wrote:
> > AIXI says that a perfect solution is not computable. However, a very
> > general principle of both scientific research and machine learning is to
> > favor simple hypotheses over complex ones. AIXI justifies these practices
> > in
AIXI says that a perfect solution is not computable. However, a very
general principle of both scientific research and machine learning is to
favor simple hypotheses over complex ones. AIXI justifies these practices
in a formal way. It also says we can stop looking for a universal
solution, whi
A reminder to all in the San Fran Bay area ... a hands-on workshop on
OpenCog hosted by Ben Goertzel, Joel Pitt and David Hart will be held
Sunday. Details are here:
http://opencog.org/wiki/CogDev2008
There is no cost except your sanity.
-- Ben G
--
Ben Goertzel, PhD
CEO, Novamente LLC and
On Sat, Oct 25, 2008 at 1:17 PM, Russell Wallace
<[EMAIL PROTECTED]> wrote:
> On Sat, Oct 25, 2008 at 9:57 AM, Vladimir Nesov <[EMAIL PROTECTED]> wrote:
>> Note that people are working on this specific technical problem for 30
>> years, (see the scary amount of work by Cousot's lab,
>> http://www.d
On Sat, Oct 25, 2008 at 9:57 AM, Vladimir Nesov <[EMAIL PROTECTED]> wrote:
> Note that people are working on this specific technical problem for 30
> years, (see the scary amount of work by Cousot's lab,
> http://www.di.ens.fr/~cousot/COUSOTpapers/ ), and they are still
> tackling fixed invariants,
On Sat, Oct 25, 2008 at 12:40 PM, Russell Wallace
<[EMAIL PROTECTED]> wrote:
>
>> What I see as potential way of AI in program analysis is cracking
>> abstract interpretation, automatically inventing invariants and
>> proving that they hold, using these invariants to interface between
>> results of
On Sat, Oct 25, 2008 at 9:29 AM, Vladimir Nesov <[EMAIL PROTECTED]> wrote:
> There are systems that do just that, constructing models of a program
> and representing conditions of absence of a bug as huge formulas. They
> work with various limitations, theorem-prover based systems using
> counterex
On Sat, Oct 25, 2008 at 3:17 AM, Russell Wallace
<[EMAIL PROTECTED]> wrote:
> On Fri, Oct 24, 2008 at 7:42 PM, Vladimir Nesov <[EMAIL PROTECTED]> wrote:
>> This general sentiment doesn't help if I don't know what to do specifically.
>
> Well, given a C/C++ program that does have buffer overrun or s
41 matches
Mail list logo