@Jan Klauck, @Mike Archibold, @AT
@Jan Klauck
>> A "philosophy first" approach means simply "first define the kind of
>> problems an AGI must solve." Or: "first define what intelligence is
>Much knowledge about the subject or good engineering design is often found
>(accidentally) by the constructive and experimental approach of building
>something and seeing what it does.
>Philosophizing about it most probably won't do the job, it can't define
>the problem nor specify the goal in a way that the rest would be just
>implementation. Theoretical grounding and philosophical orientation are
>useful, but "philo first" means armchairing and avoiding the dirty work.
That's right, some of the reasons are that human memory and mind have too
small capacity - one cannot keep in her mind too much, neither she can
imagine in details, without using some external tools, real "unwinded"
operation of complex systems and their by-effects.
@Mike Archibold:
>I think what is meant by "simple" in the agi context surrounds the
>issue of designating some starting point that is manageable. There
>are so many possibilities for a START out there: math-first start
>(develop the correct formulas), logic-first, psychology-first,
>neuroscience-first, physics-first, old-AI first ("we just need the
>right algorithms").... my book argues for a philosophy-first starting
>point (I finally got it sent to the printer!).
IMO - all should go together - there's no first or second, all in parallel,
all should explain and justify each other. If somebody can't manage it -
sorry, that's apparently too big bite for her mouth. And it is, for most
AGI researchers, they are not multi-disciplinary and inter-disciplinary
enough.
As some philosopher schools said a century ago - theory without practice
(constant grounding) turns into bullshit and schizophrenical nonsense, and
practice without theory is doomed not to scale, to be stay something like a
handicraft.
Inductive and deductive reasoning should go together, the abstract directs
the lower perceptions, but the lower perceptions should build the
abstraction, they approve and disprove them.
Also meaningful philosophy is induced, supported or dismissed by all those
other sciences and their specific results.
@AT
AT> Of course all kinds of simple learning have been tried for decades
AT> But what may have been missed by a collective IQ of a million or a
billion?
Integration is missing, the researchers were not generally intelligent, so
they couldn't see the whole picture and integrate the solution which is
obvious and explained, but the obvious explanation is spread in too many
research fields and data records. Human working memory and any human kind
of memory is very tiny, attention span - as well.
(In addition, absurdly, decent virtual worlds are still being developed.)
That tiny human capacity is a fundamental problem of AGI and any research.
Problems that could be solved in one or a few passes are solved in 10000000
passes, consisting of 3849384938493 sub-passes each. No matter what
computers we do have now, human's memory is still limited like the 64KB of
the classic 8-bit microcomputers - in fact it's very much worse, - and the
speed is like of the Apple][ floppy drive - actually much worse.
Unfortunately everything eventually goes there for evaluation, that makes
the progress slow and sometimes random, or just causes "out of memory",
until the external tools reach to a state when they can reduce the input
down to some digestible tiny amount of data model.
Bootstrapping is critical.
--
===* Todor "Tosh" Arnaudov ===*
*
.... Twenkid Research:* http://research.twenkid.com
.... *Author of the world first University courses in AGI (2010, 2011)*:
http://artificial-mind.blogspot.com/2010/04/universal-artificial-intelligence.html
*.... Todor Arnaudov's Researches Blog**: *
http://artificial-mind.blogspot.com
*
*
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com
