Hi,
It follows that the AIXItl algoritm applied to friendliness would be
effectively more friendly than any other time t and space bounded
agent.
Personally I find that satisfying in the sense that once
"compassion", "growth" and "choice" or the classical "friendliness"
has been defined an opti
Hi again,
On 9/11/06, Ben Goertzel <[EMAIL PROTECTED]> wrote:
> It follows that the AIXItl algoritm applied to friendliness would be
> effectively more friendly than any other time t and space l bounded
> agent.
Yes, but the problem is that AIXItl, in order to run effectively,
requires unpractic
Meanwhile, the biologists continue making excellent, steady progress
toward understanding how the curse of aging operates
http://www.hhmi.org//news/morrison20060906.html
"
A single molecular switch plays a central role in inducing stem cells
in the brain, pancreas, and blood to lose function
Hi,
On Ben's essay: Ben is arguing that due to incomputable complexity
'friendliness' can only be guranteed under unsatisfactory narrow
circumstances. Independent of one agrees or not, it would follow that
if this is the case then substituting friendliness with one or all of
the alternative goal
Since we assume that AIXItl is effectively better at achieving its
goal than any other agent with the same space and time resource
limitations the specific values for t and l do not matter. Do they?
> In short: it's some pretty math with some conceptual evocativeness,
> but not of any pragmatic v
On 9/11/06, Stefan Pernar <[EMAIL PROTECTED]> wrote:
Assuming that AIXItl be effectively better at achieving its goal thanany other agent with thesame space and time resource limitations (t,l) would still make it thealgorithm of choice no matter how computationally intense, as any
other system give
This discussion on how to guarantee that AI be friendly seems to presume that a
super AI and human brains will be separate entities. I believe this will not
be the case, and the issue will be moot.
Humans would like to be smarter: to be able to think faster, learn faster,
communicate faster, a
On 9/11/06, Matt Mahoney <[EMAIL PROTECTED]> wrote:
This
discussion on how to guarantee that AI be friendly seems to presume
that a super AI and human brains will be separate entities. I believe
this will not be the case, and the issue will be moot.
I was actually planning a followup to my "negati
From: Ben Goertzel <[EMAIL PROTECTED]>
> http://www.hhmi.org//news/morrison20060906.html
Unfortunately, limiting the number of cell divisions is necessary to prevent
cancer. We also need a way to reduce the mutation rate.
-- Matt Mahoney, [EMAIL PROTECTED]
- Original Message
To: sing
The subtle question raised by the AI/human fusion approach is: Once
you become something vastly more intelligent and general than "human",
in what sense are you still "you" ... ?? In what sense have you
simply killed yourself slowly (or not so slowly) and replaced yourself
with something cleverer
On 9/11/06, Ben Goertzel <[EMAIL PROTECTED]> wrote:
The subtle question raised by the AI/human fusion approach is: Onceyou become something vastly more intelligent and general than "human",in what sense are you still "you" ... ?? In what sense have you
simply killed yourself slowly (or not so slow
On 9/11/06, Shane Legg <[EMAIL PROTECTED]> wrote:
(...)
I'm no longer the child who used to play with the kids across the
road in New Zealand in the late 1970's. I have some memories of
being that person, but that's it. I don't see this as fundamentally
different, especially if it is allowed to
Technologically, AI is far far easier than uploading. So, AI will
come first, and we will have to build AI that is reliably nice to us,
or suffer the consequences.
It's not "controlling an entity smarter than you" if it was built from
scratch to be nice, and continues to be nice of its own accor
On 9/11/06, Michael Anissimov <[EMAIL PROTECTED]> wrote:
Technologically, AI is far far easier than uploading. So, AI will
come first, and we will have to build AI that is reliably nice to us,
or suffer the consequences.
(...)
I am not so sure that AI is easier than uploading.
Surely upload n
Joel,
You make some good points that make me think deeper about this, but I'm
still locked into a non-intentional KC. I'm going to try to explain this in
light of my defnition of knowledge creation, which is different from others
out there today.
Per one of my last notes, knowledge is a str
Bruce,
Thanks for taking the time for that detailed response.
I'll respond to your views on AKC once I've had a read through
your book chapter.
In the mean time:
Some of the technical discussion in this forum is tough for me to follow,
but it seems to me that this is where most of the discussi
Lucio wrote:
In order to produce strong AI, though, we need to understand the mind
from low to high levels, or understand the processes that make high
levels emerge from low ones. That seems a much tortuous scientific
path, one that cannot be achieved by conventional and comparatively
predictabl
On 9/12/06, Russell Wallace <[EMAIL PROTECTED]> wrote:
On 9/11/06, Stefan Pernar <[EMAIL PROTECTED]> wrote:
> Assuming that AIXItl be effectively better at achieving its goal than
> any other agent with the
> same space and time resource limitations (t,l) would still make it the
> algorithm of c
On 9/12/06, Stefan Pernar <[EMAIL PROTECTED]> wrote:
An interesting question for me now is when the comparativeinefficencies of the AIXItl algorythms will become irrelevant due toever decreasing cost and availability of computational resources.
It won't. Basically AIXI and suchlike work by exhaust
Hi,
Just for kicks - let's assume that AIXItl yields 1% more intelligent
results when provided 10^6 times the computational resources when
compared to another algorythm X. Let's further assume that today the
cost asscociated with X for reaching a benefit of 1 will be 1 compared
to a cost of 10^6
On 9/12/06, Ben Goertzel <[EMAIL PROTECTED]> wrote:
In this scenario it will be
computationally cheaper to apply the AIXItl in less than 20 years.
But this is nowhere near reality -- the level of inefficiency of
AIXItl is such that it will never be usable within the physical
universe, unless c
Thanks Ben, Russel et al for being so patient with me ;-) To
summarize: AIXItl's inefficiencies are so large and the additional
benefit it provides is so small that it will likely never be a logical
choice over other more efficient, less optimal algorithms.
Stefan
The additional benefit it *wou
22 matches
Mail list logo