--- Jef Allbright <[EMAIL PROTECTED]> wrote:
> On 7/1/07, Stathis Papaioannou <[EMAIL PROTECTED]>
> wrote:
>
> > If its top level goal is to allow its other goals
> to vary randomly,
> > then evolution will favour those AI's which decide
> to spread and
> > multiply, perhaps consuming humans in
On 7/1/07, Stathis Papaioannou <[EMAIL PROTECTED]> wrote:
If its top level goal is to allow its other goals to vary randomly,
then evolution will favour those AI's which decide to spread and
multiply, perhaps consuming humans in the process. Building an AI like
this would be like building a bomb
--- Stathis Papaioannou <[EMAIL PROTECTED]> wrote:
> On 02/07/07, Tom McCabe <[EMAIL PROTECTED]>
> wrote:
>
> > > For
> > > example, if it has as its most important goal
> > > obeying the commands of
> > > humans, that's what it will do.
> >
> > Yup. For example, if a human said "I want a
> bana
On 02/07/07, Tom McCabe <[EMAIL PROTECTED]> wrote:
> For
> example, if it has as its most important goal
> obeying the commands of
> humans, that's what it will do.
Yup. For example, if a human said "I want a banana",
the fastest way for the AGI to get the human banana
may be to detonate a kilo
The problem isn't that the AGI will violate its
original goals; it's that the AGI will eventually do
something that will destroy something really important
in such a way as to satisfy all of its constraints. By
setting constraints on the AGI, you're trying to think
of everything bad the AGI might p
--- Stathis Papaioannou <[EMAIL PROTECTED]> wrote:
> On 02/07/07, Tom McCabe <[EMAIL PROTECTED]>
> wrote:
>
> > The AGI doesn't care what any human, human
> committee,
> > or human government thinks; it simply follows its
> own
> > internal rules.
>
> Sure, but its internal rules and goals migh
On 02/07/07, Tom McCabe <[EMAIL PROTECTED]> wrote:
The AGI doesn't care what any human, human committee,
or human government thinks; it simply follows its own
internal rules.
Sure, but its internal rules and goals might be specified in such a
way as to make it refrain from acting in a particul
On 02/07/07, Tom McCabe <[EMAIL PROTECTED]> wrote:
But killing someone and then beating them on the
chessboard due to the lack of opposition does count as
winning under the formal rules of chess, since there's
nothing in the rules of chess about killing the
opponent. The rule "don't kill, strangl
On 02/07/07, Eugen Leitl <[EMAIL PROTECTED]> wrote:
> But in the final analysis, the AI would be able to be implemented as
> code in a general purpose language on a general purpose computer with
Absolutely not. Possibly, something like a silicon compiler
with billions to trillions asynchronous
BillK wrote:
> On 7/1/07, Tom McCabe wrote:
>> These rules exist only in your head. They aren't
>> written down anywhere, and they will not be
>> transferred via osmosis into the AGI.
> They *are* written down.
> I just quoted from the FIDE laws of chess.
> And they would be given to the AGI alon
--- BillK <[EMAIL PROTECTED]> wrote:
> On 7/1/07, Tom McCabe wrote:
> >
> > These rules exist only in your head. They aren't
> > written down anywhere, and they will not be
> > transferred via osmosis into the AGI.
> >
>
> They *are* written down.
> I just quoted from the FIDE laws of chess.
> A
On 7/1/07, Tom McCabe wrote:
These rules exist only in your head. They aren't
written down anywhere, and they will not be
transferred via osmosis into the AGI.
They *are* written down.
I just quoted from the FIDE laws of chess.
And they would be given to the AGI along with the rest of the rul
--- BillK <[EMAIL PROTECTED]> wrote:
> On 7/1/07, Tom McCabe wrote:
> > The constraints of "don't shoot the opponent"
> aren't
> > written into the formal rules of chess; they exist
> > only in your mind. If you claim otherwise, please
> give
> > me one chess tutorial that explicitly says "don't
On 7/1/07, Tom McCabe wrote:
The constraints of "don't shoot the opponent" aren't
written into the formal rules of chess; they exist
only in your mind. If you claim otherwise, please give
me one chess tutorial that explicitly says "don't
shoot the opponent".
This is just silly.
All competiti
--- Eugen Leitl <[EMAIL PROTECTED]> wrote:
> On Sun, Jul 01, 2007 at 01:51:26PM -0700, Tom McCabe
> wrote:
>
> > All of this applies only to implosion-type
> devices,
> > which are far more complicated and tricky to pull
> off
> > than gun-type devices, and which are therefore
> > unlikely to be
--- Eugen Leitl <[EMAIL PROTECTED]> wrote:
> On Sun, Jul 01, 2007 at 08:56:24PM +1000, Stathis
> Papaioannou wrote:
>
> > But the constraints of the problem are no less a
> legitimate part of
>
> We're all solving the same problem: sustainable
> self-replication
> long-term.
What does self-rep
The constraints of "don't shoot the opponent" aren't
written into the formal rules of chess; they exist
only in your mind. If you claim otherwise, please give
me one chess tutorial that explicitly says "don't
shoot the opponent".
- Tom
--- Stathis Papaioannou <[EMAIL PROTECTED]> wrote:
> On 01/
--- Eugen Leitl <[EMAIL PROTECTED]> wrote:
> On Sun, Jul 01, 2007 at 12:44:09AM -0700, Tom McCabe
> wrote:
>
> > > They also need knowledge, which is still largely
> > > secret.
> >
> > Knowledge of *what*? How to build a crude gun to
> fire
> > one block of cast metal into another block of cas
Samantha Atkins wrote:
Sergey A. Novitsky wrote:
Dear all,
...
o Be deprived of free will or be given limited free will (if
such a concept is applicable to AI).
See above, no effective means of control.
- samantha
There is *one* effective means of control:
An AI wil
Peter Voss wrote:
Just for the record: To put it mildly, not everyone is 'Absolutely' sure
that AGI can't be implemented on Bill's computer.
In fact, some of us are pretty certain that that (a) current hardware is
adequate, and (b) AGI software will be with us in (much) less than 10 years.
Some
--- Eugen Leitl <[EMAIL PROTECTED]> wrote:
> On Sun, Jul 01, 2007 at 07:01:07PM +1000, Stathis
> Papaioannou wrote:
>
> > What sort of technical information, exactly, is
> still secret after 50 years?
>
> The precise blueprint for a working device. Not a
> crude gun assembler,
> the full implos
But killing someone and then beating them on the
chessboard due to the lack of opposition does count as
winning under the formal rules of chess, since there's
nothing in the rules of chess about killing the
opponent. The rule "don't kill, strangle, drug, maim,
injure, or otherwise physically hurt"
Samantha Atkins wrote:
Charles D Hixson wrote:
Stathis Papaioannou wrote:
Available computing power doesn't yet match that of the human brain,
but I see your point, software (in general) isn't getting better
nearly as quickly as hardware is getting better.
Well, not at the personally accessib
On Sun, Jul 01, 2007 at 01:51:26PM -0700, Tom McCabe wrote:
> All of this applies only to implosion-type devices,
> which are far more complicated and tricky to pull off
> than gun-type devices, and which are therefore
> unlikely to be used.
We're arguing something tedious, and meaningless.
The
--- Eugen Leitl <[EMAIL PROTECTED]> wrote:
> On Sun, Jul 01, 2007 at 12:45:20AM -0700, Tom McCabe
> wrote:
>
> > Do you have any actual evidence for this? History
> has
> > shown that numbers made up on the spot with no
> > experimental verification whatsoever don't work
> well.
>
> You need 10
--- Eugen Leitl <[EMAIL PROTECTED]> wrote:
> On Sun, Jul 01, 2007 at 12:47:27AM -0700, Tom McCabe
> wrote:
>
> > Because an AGI is an entirely different kind of
> thing
> > from evolution. AGI doesn't have to care what
>
> If it's being created by an evolutionary context
What does that even me
Eugen Leitl wrote:
> On Sun, Jul 01, 2007 at 11:11:06PM +1000, Stathis Papaioannou wrote:
>> But in the final analysis, the AI would be able to be implemented as
>> code in a general purpose language on a general purpose computer with
> Absolutely not. Possibly, something like a silicon compiler
On Sun, Jul 01, 2007 at 10:08:53AM -0700, Peter Voss wrote:
> Just for the record: To put it mildly, not everyone is 'Absolutely' sure
> that AGI can't be implemented on Bill's computer.
What is Bill's computer? A 3 PFlop Blue Gene/P? A Core 2 duo box from Dell?
> In fact, some of us are pretty
Just for the record: To put it mildly, not everyone is 'Absolutely' sure
that AGI can't be implemented on Bill's computer.
In fact, some of us are pretty certain that that (a) current hardware is
adequate, and (b) AGI software will be with us in (much) less than 10 years.
Some people may be very,
On Sun, Jul 01, 2007 at 11:11:06PM +1000, Stathis Papaioannou wrote:
> But in the final analysis, the AI would be able to be implemented as
> code in a general purpose language on a general purpose computer with
Absolutely not. Possibly, something like a silicon compiler
with billions to trillion
Samantha Atkins wrote:
> Alan Grimes wrote:
>>> Available computing power doesn't yet match that of the human brain,
>>> but I see your point,
>> What makes you so sure of that?
> It has been computed countless times here and elsewhere that I am sure
> you are aware of so why do you ask?
http://
On 01/07/07, Eugen Leitl <[EMAIL PROTECTED]> wrote:
I'm not sure all-purpose hardware would be suitable [for AI]. It depends very
much
upon which computing paradigm is dominant by that time (40-50 years
away from now). Judging from the past, there might be not that much
progress there.
But in
On Sun, Jul 01, 2007 at 08:56:24PM +1000, Stathis Papaioannou wrote:
> But the constraints of the problem are no less a legitimate part of
We're all solving the same problem: sustainable self-replication
long-term. Artificial systems so far are only instruments (though
they do evolve in the human
On 01/07/07, Eugen Leitl <[EMAIL PROTECTED]> wrote:
> that disabling your opponent would be helpful, it's because the
> problem it is applying its intelligence to is winning according to the
> formal rules of chess. Winning at any cost might look like the same
> problem to us vague humans, but i
On Sun, Jul 01, 2007 at 12:44:09AM -0700, Tom McCabe wrote:
> > They also need knowledge, which is still largely
> > secret.
>
> Knowledge of *what*? How to build a crude gun to fire
> one block of cast metal into another block of cast
> metal?
How about gas centrifuges, materials resistant to
U
On 01/07/07, Tom McCabe <[EMAIL PROTECTED]> wrote:
> If its goal is "achieve x using whatever means
> necessary" and x is
> "win at chess using only the formal rules of chess",
> then it would
> fail if it won by using some means extraneous to the
> formal rules of
> chess, just as surely as it
On Sun, Jul 01, 2007 at 07:01:07PM +1000, Stathis Papaioannou wrote:
> What sort of technical information, exactly, is still secret after 50 years?
The precise blueprint for a working device. Not a crude gun assembler,
the full implosion assembly mounty. The HE lens geometries, the timing,
the me
On Sun, Jul 01, 2007 at 12:45:20AM -0700, Tom McCabe wrote:
> Do you have any actual evidence for this? History has
> shown that numbers made up on the spot with no
> experimental verification whatsoever don't work well.
You need 10^17 bits and 10^23 ops to more or less accurately
represent and t
On 01/07/07, Eugen Leitl <[EMAIL PROTECTED]> wrote:
They also need knowledge, which is still largely secret. The
number of people who could construct a crude working nuclear weapon
is below 1% (possibly, way below 1%), and an advanced fusion weapon
in ppm range.
What sort of technical informat
On Sun, Jul 01, 2007 at 12:47:27AM -0700, Tom McCabe wrote:
> Because an AGI is an entirely different kind of thing
> from evolution. AGI doesn't have to care what
If it's being created by an evolutionary context and
is competing with likewise it is precisely evolution,
only with a giant fitness
Because an AGI is an entirely different kind of thing
from evolution. AGI doesn't have to care what
evolution is or how it works; there's no constraint on
it whatsoever to act like evolution does. Evolution is
actually nicer than most AGIs, because evolution is
constrained by the need to have one v
--- Eugen Leitl <[EMAIL PROTECTED]> wrote:
> On Sat, Jun 30, 2007 at 10:11:20PM -0400, Alan
> Grimes wrote:
>
> > =\
> > For the last several years, the limiting factor
> has absolutely not been
> > hardware.
>
> How many years? How much OPS, aggregated network and
> memory bandwidth?
> What is
--- Eugen Leitl <[EMAIL PROTECTED]> wrote:
> On Sun, Jul 01, 2007 at 10:24:17AM +1000, Stathis
> Papaioannou wrote:
>
> > Nuclear weapons need a lot of capital and
> resources to construct,
>
> They also need knowledge, which is still largely
> secret.
Knowledge of *what*? How to build a crude
On Sun, Jul 01, 2007 at 11:35:21AM +1000, Stathis Papaioannou wrote:
> Why do you assume that "win at any cost" is the default around which
> you need to work?
Because these are the rules of the game for a few GYrs. Why do you
assume that these have changed?
-
This list is sponsored by AGIRI
On Sat, Jun 30, 2007 at 10:11:20PM -0400, Alan Grimes wrote:
> =\
> For the last several years, the limiting factor has absolutely not been
> hardware.
How many years? How much OPS, aggregated network and memory bandwidth?
What is your evidence for your claim?
> How much hardware do you claim y
On Sun, Jul 01, 2007 at 10:37:43AM +1000, Stathis Papaioannou wrote:
> But Deep Blue wouldn't try to poison Kasparov in order to win the
> game. This isn't because it isn't intelligent enough to figure out
Yes, it is precisely because the system is not intelligent enough.
> that disabling your o
On Sun, Jul 01, 2007 at 10:24:17AM +1000, Stathis Papaioannou wrote:
> Nuclear weapons need a lot of capital and resources to construct,
They also need knowledge, which is still largely secret. The
number of people who could construct a crude working nuclear weapon
is below 1% (possibly, way belo
47 matches
Mail list logo
