Monday, August 4, 2014

Where calculations begin, comprehension ceases* - on understanding and superintelligent AGI. Todor's comment in "Strong Artificial Intelligence" at Google+

Comment on the thread: https://plus.google.com/u/0/+DuncanMurray_AU/posts/NeuVTZonghR?cfem=1

Todor Arnaudov:

Yes, in too simple situations or ones where the possibilities for action are physically limited, by other unalterable superior forces, a superior intelligence is not obviously useful. A genius in an empty solid room with no doors, no windows, no holes, and no tools wouldn't invent better ways to escape than any normal guy or a cockroach.

Also - yeah, most people, having current biology, don't need technology beyond "baking a cake" (food, drinking, sex, entertaining (legal sensory-taken drugs such as TV), social approval) - people don't really care about technology, and hopefully there were ones who did, otherwise we would have still be dying in the caves.

Average people wouldn't see it as useful for the sake of its cleverness, it's not directly "practical" for an animal (sex, food, ...), it may actually be "anti-practical", because it will show humans that they are dumb,  that "there's no need for human work, since it can be done/synthesized in 0.1 secs by one machine" - and many may go mad about that; many of their recent values will lose their significance and it will become obvious that they are not "magical".

....

I agree that one of the crucial fields for the AGI is biology. There are a few other important fields which will be such in the beginning, but they will do because they are easy and actually will be "anihilated" and "eaten up" immediately, since they are obvious and trivial, they don't even need enormous computing power for today's standards.

.....

It is true also that the "normal" people usually don't need genii (including super smart AGI),  and a genius, as explained by Schopenhauer some 200 years ago is usually useless for his contemporaries, because they can't understand him; they see him as a "crackpot", while a genius could see the other people as a bunch of silly "idiots"  (and would be correct). As of the Matt's example about Einstein as a possible crackpot for the "most people" - do you bear in mind that most people have "learning disabilities", can't learn even basic Calculus, can't learn to draw, to juggle, to play musical instruments, to write a decent novel, can barely learn a foreign language or two - and would talk terribly. They are "retards" in most of the creative fields, if measured honestly and compared to the "talented" ones. Of course Einstein was a genius in theoretical physics, not an ingenius car driver, as Michael Schumacher is an ingenious car driver ("practical real-time physicist with high-speed sensory-motor physical processing, sensory-muscular reaction with a high precision, high-precision real-time trajectory prediction etc.), not an ingenious theoretical physicist (not real time, higher generality data, ...) - there's nothing hard in measuring or recognising this.

......

So why don't we turn the glass on the other side - from Einstein's point of view most of the people are idiots and retards, they can't see and cannot learn what was obvious and trivial to him - while he completely understood and could match their inferior understanding of theoretical physics (of which most people don't even have a clue, above some obvious elements of classical mechanics, without formalization).

.......

If someone is capable to recreate/understand/explain/trace/unveil/expand/repeat... what you can do, if he is capable to emulate you; but you're not capable to emulate him - or if he does it faster, deeper, longer, more sustainable ... - he's better (smarter, more clever, superior, ...) than you in that domain. Of course for low level definition such a natural language one is too general.
 ....

In general, the above is related to the amount of working memory (and access for particular types of sensory motor memory), in a wider sense that the term from the cognitive science, which even in its narrow sense is proven and obviously correlated with the G-factor.

....

 Most people cannot understand at all some topics, can't have a clue and can't learn them (besides to memorize some fragments , terms, concepts, operations etc. by heart that they can recite, perform, apply etc., but cannot continue, extend, optimize), namely because they are unable to put the problem in their mind.

.....

The ones who do not understand see random unconnected fragments and cannot follow the thought process,  besides on the "tracks" that are memorized as sequences to be "replayed". They can't see the links, do not understand the *PHYSICS* (the causality), the "intentionallity", the "philosophy", the idea (in Schopenhauer's sense), and can't create an intuitive view of the domain/question ("intuitive" in the sense of Kant and Schopenhauer ). So that's why they don't see anything meaningful or notice only "miracles", "radical novelty" or other nonsenses (like in art for the people who do not understand it and can't learn it) and the only way for them to pretend to understand such topics is to see some by-effects, some results - which however is not understanding of the topic, but "noticing" a detail, or an "approval" in other simpler and more trivial terms/domains/context); or if they follow straightforward logical chains "if-...then... therefore", possibly long,  but be allowed to forget the past steps, so that the process is simple enough to have small amount of data within the single steps. At the end, they could reach at the conclusion and check that it was "correct", but otherwise they could not see the whole picture at once, and that's not real understanding/comprehension.

....

 As Schopenhauer has said some 200 years ago for humans for such cases: "Where calculations begin, comprehension ceases" - in "On the fourfold root of the principle of sufficient reason", which ineed is one of the true early seminal AGI monographs, among Kant's critiques and all of the Schopenhauer's works. That was some 180 years before the often cited "seminal paper" of "Symbol Grounding" (1990) and 170 before the Searle's pathetic AGI "disproval" with his "Chinese Room experiment" (1980). Humans suffer from the "Chinese Room" or more properly - Schopenhauer's "words instead of thoughts" issue themselves, they did suffer very hardly centuries ago (in the philosophical sophisms). Reciting memories, reordering words without taking care of their meaning and connections, without holding in mind these connections; doing blind calculations, without understanding why, what, for what etc. - that's what most people do anyway. And that is not understanding. In one point of view, even the brightest genius doesn't really understand the technology - he cannot hold the processing inside it in his mind even in the slightest detail, besides extremely tiny or general fragments at a time.

.....

We would not be able to understand thoroughly and "really" the thought process of an AGI that is too much smarter than us, as well. Only the principles or fragmentary and it may look like "magic".
 However as of such "general" understanding of principles, even laymen "understand"/are capable to *say* something about some of the most general and easy to be uttered principles behind computer science - "digital", "1s and 0s", "flip-flops", "has memory cells where you store data", "it does what you tell it to do" etc., while at the same time he may be unable to code even a trivial system of 40 lines of code or understand how a simple adder etc. really works. So one may question whether he does understand what he's talking about, or he's just a "talking machine", an advanced "speech synthesizer".


*"Where calculation begins, comprehension ceases" - a Schopenhauer's thought from "On the fourfold root of the principle of sufficient reason", year 1813 (?+editions). From the English translation at https://archive.org/stream/onthefourfoldroo00schouoft/onthefourfoldroo00schouoft_djvu.txt

Translation in English by William T Fee, 1903.

Pages 90-91:

Also: "Calculation conveys no comprehension ... [it is] of merely practical, not theoretical value ... ... deals exclusively with abstract concepttions of magnitude, whose mutual relations ..."




 To calculate therefore, is not to understand, and,
in itself, calculation conveys no comprehension of things.
Calculation deals exclusively with abstract conceptions of
magnitudes, whose mutual relations it determines. By it
we never attain the slightest comprehension of a physical
process, for this requires intuitive comprehension of
space-relations, by means of which causes take effect.


FIKST CLASS OF OBJECTS FOR THE SUBJECT. 91

Calculations have merely practical, not theoretical, value.
It may even be said that where calculation begins, compre-
hension ceases ; for a brain occupied with numbers is, as
long as it calculates, entirely estranged from the causal
connection in physical processes, being engrossed in purely
abstract, numerical conceptions. The result, however, only
shows us how much, never what.

p.17


And if Neo-Spinozans (Schellingites, Hegelians,
&c.), with whom words are wont to pass for thoughts


p.44

Germans are accustomed to content themselves
with words instead of thoughts
. Do we not train them
to it from their cradle? Only look at Hegelianism!
What is it but empty, hollow, nauseous twaddle!


Arthur Schopenhauer, 1813 (Bold - T.A.)


No comments :