<< ctxmodel.net

> Another question is how close do you think mankind is to
> true arificial intelligence.

That depends on a definition.
I think that there're already no theoretical problems in android production, except for high software development costs and inavailability of any applications for said androids (people are much cheaper anyway).

Also I think that making of a human-like artifical intelligence requires an android hardware, because its an integral thing and even tasks like natural language translation need a huge database of information which can be collected only by something human-like living in a society.

So any practical AI-related tasks like recognition or decision-making are solved (and would be) in a way unrelated to self-awareness and real intelligence (and btw the same models like used in statistical compressors are as well necessary for any of these tasks), and we probably can only expect something human-like to be created by entertainment industry or some hobbyists.

2008-12-06 22:41:36 chornobyl       >
Yesterday i talk with one friend of mine, and he where highly anticipated against AI in the form of decision making. He where afraid of machines rebel against humans. Actually if i where machine i would certainly do so 8].
My point of the whole AI idea, when we will make it working we'll get yet another instrument to explore universe (like microscope, telescope..). It will help us understand things which we cannot now because of memory and/or speed limitations of our mind.
2008-12-07 09:08:50 Shelwien        >
Dunno how being a machine is related to rebelling against humans.
I think you're perfectly able to do that even as what you are now :)
So just go and conquer the world ;)
Anyway, humans are produced with a higher level technology than
currently available... materials are very cheap, and still they're
much more durable than any factory-made machines or electronics.
So you have a better chance than being a machine ;)
And considering software aspects, its no different. Even simple
software reverse-engineering is hardly possible without external
access to hardware, especially the display (there're no perfect emulators),
and the ability to read (and OCR!) human languages. That's beside
the fact that software decision-making is still comparable
with human's in speed (commonly slower), as most CPUs don't yet
have enough cores to compete with hardware NN in parallel execution.
So tell your friend that he's safe at least for 10 years.

And as to "instruments", its just a matter of direct interface.
Check out something like
2008-12-07 19:11:42 Shelwien        >
>--- Toffer       07.12.2008 18:11
well, that's too philosophic, i think
and i bet it'll take more than 10 years

>--- Shelwien     07.12.2008 18:11
not sure
AI tasks are quite practical like text translation
or automated software analysis, including reverse engineering
also various less known areas like image/video description (textual).
and i think that considering only the theory and algorithms,
its not a problem already
but getting the content for knowledge database is problematic

>--- Toffer       07.12.2008 18:15
i'm really unsure about that
another thing is that a brain works much more different from a cpu

>--- Shelwien     07.12.2008 18:16
that's unrelated
though its not a problem to build or simulate a NN

>--- Toffer       07.12.2008 18:16
but the nn models are far from that what a biological neuron does

>--- Shelwien     07.12.2008 18:17
there're biological neuron models as well already
complex but whatever
it just won't solve anything
well, unless we'd manage to gradually replace the brain of some person with
neuron emulators

>--- Toffer       07.12.2008 18:19
and i still think it's science fiction

>--- Shelwien     07.12.2008 18:19
not really
just too expensive
but it'd move pretty fast with nanotechnology
however as i said, its unrelated
software solution of AI tasks is what matters
and that can't be done by even a perfect brain model
if that's a model without content
and i think that content is what matters
if somebody manages to record what a person sees and feels
in first 20 years of life i think that'd be enough to
construct an AI with known methods
using any elements to approximate the necessary models

>--- Toffer       07.12.2008 18:23
this is surely another problem
rather complex, too

>--- Shelwien     07.12.2008 18:23
so imho its as i said in that post
until somebody builds a quality android, we won't move anywhere
and such a thing is pretty unusable, except for entertainment purposes
but very expensive at the same time
that's why we don't have AI yet ;)

>--- Toffer       07.12.2008 18:25
so we need more money for the entertaiment industry :)
2013-07-27 19:17:14                 >
2014-11-26 14:13:48                 >
2015-01-11 11:32:51                 >

Write a comment: