http://shilinc.files.wordpress.com/2011/06/e99d9ee5b9bfe5bbb6e7bb9fe8aea1e58a9be5ada6e4b88ee5ae8ce585a8e5bc80e694bee7b3bbe7bb9fe79a84e7bb9fe8aea1e58886e5b883.pdf
狗译原文(http://thedailyomnivore.net/2013/01/23/moravecs-paradox/):
Moravec的矛盾是人工智能和机器人技术研究人员发现,传统的假设相反,高层次的推理只需要很少的计算,但低级别的感觉运动技能需要巨大的计算资源。汉斯·莫拉维克,罗德尼·布鲁克斯,马文·明斯基在20世纪80年代明确提出的原则。莫拉维克写道,“这是比较容易使计算机具有成人级的性能智力测验或玩跳棋,很难或根本不可能给他们一岁的时候,它涉及到感知和流动性的技能。”
语言学家和认知科学家史蒂芬·平克认为这是最重要的发现人工智能研究人员发现。在他的著作“语言本能”,他写道:“的AI研究35年的主要教训是,困难的问题是很容易的,容易的问题是很难。的心理能力,一个4岁,我们理所当然的 - 识别脸部,拿起铅笔,在房间内行走,回答一个问题 - 事实上,解决一些最困难的工程问题,你的...作为新一代的智能设备的出现,这将是股票分析师和石油化工工程师和假释委员会成员被机器所取代的危险。的园丁,接待和厨师在他们的工作是安全的,几十年来“。马文·明斯基强调,最困难的人的技能进行逆向工程,是无意识的。“在一般情况下,我们至少知道什么我们的脑海中做得最好的,”他写道,并补充说:“我们知道简单的过程,不能很好的工作较复杂的工作完美。”
一个可能的解释的悖论,所提供的莫拉维克,是基于进化。实施生物,人类所有的技能使用自然选择的过程中,机械设计。在他们的进化过程中,自然选择倾向于保留设计的改进和优化。年长的一个技巧是,自然选择有更多的时间,提高了设计。抽象思维,只是最近才开发的,因此,我们不应该指望它的实现是特别有效的。莫拉维克写道:“大,人的大脑高度进化的感觉和运动部分的编码是一个10亿年的关于世界的性质和如何生存在它的经验。故意的过程中,我们称之为推理,我相信,人类的思想,最薄的单板有效的,不仅是因为它的支持,这大年纪了,更强大,但通常是无意识的,感觉知识。我们都是巨大的奥运选手在知觉和运动的地区,那么好,我们做的难看起来很容易。但是,抽象思维,是一种新的伎俩,也许不到10万年的历史。我们还没有掌握它。这是不是所有的内在的困难,它只是似乎,所以当我们做到这一点。
紧凑的方式来表达这种说法是:我们应该期待任何人的技能是技能的时间量一直在不断发展中的动物大致成正比的逆向工程的难度。最古老的人类技能在很大程度上是无意识的,所以在我们看来,是轻松。因此,我们应该期望出现的技能,毫不费力地难以反向工程,但需要努力的技能,不一定是很难工程师。一些例子技能,已被不断变化的数以百万计的年:识别脸部,移动周围的空间,判断人的动机,接球,识别语音,设置适当的目标,注意是有趣的事情;什么关系知觉,注意,可视化,运动技能,社会技能等。的技能,最近出现的一些例子:数学,工程学,人的游戏,逻辑,就是我们所说的科学。这些对我们来说是很难的,因为他们不是我们的身体和大脑的主要发展做。这些技能和技术,最近被收购,在历史的时间,并有最多几千年来加以提炼,主要是通过文化的演变。
在早期的人工智能研究,主要研究预测,他们将能够在短短的几十年间,创造思维机器。他们的乐观是因为一部分的事实,他们已经成功地编写程序逻辑,解决了代数和几何的问题,玩游戏,跳棋和国际象棋等。逻辑代数和有困难的人,被认为是智力的迹象。他们认为,(几乎)解决“硬”的问题,该如何运用自己的眼光与常识推理的'方便'问题将很快陷入到位。他们错了,原因之一是,这些问题是不容易的,但令人难以置信的困难。事实上,他们已经解决了的问题,如逻辑和代数是不相干的,因为这些机器非常容易解决的问题。
罗德尼·布鲁克斯解释说,根据早期人工智能的研究,情报是“最有特征性的事情,受过良好教育的男性科学家们发现了具有挑战性的,如象棋,象征性的整合,证明数学定理,解决复杂的词代数问题。“四五年的儿童可以做的事情,毫不费力,如视觉上区分一杯咖啡和椅子,或左右两条腿走,或从他们的卧室在客厅找到自己的方式,没有想到的是活动需要的情报。“ 这将导致布鲁克斯继续在人工智能和机器人研究的一个新的方向。他决定建立“无认知的智能机。传感和行动。这是我将建立完全离开了传统上被认为是人工智能的智慧。“ 这个新的方向,他称之为“中篇小说AI”是非常有影响力的机器人技术研究和AI。
Mind/Matter Frank Wilczek
http://cerncourier.com/cws/article/cern/32522
From BCS to the LHC
Steven Weinberg reflects on spontaneous symmetry breaking, and the connection between condensed-matter physics and particle physics, in a talk at the University of Illinois in Urbana, celebrating the 50th anniversary of the BCS theory of superconductivity.
: (Mind 6= Matter): The End of an Illusion
The distinction between Mind and Matter is embedded in everyday language and thinking,
and even more deeply in philosophy and theology. The great philosopher/theologian George
Berkeley, who famously grounded Matter in the Mind of God, summed it up in a witticism:
What is mind? No matter.
What is matter? Never mind.
Science has long found it useful to accept this duality, as a methodology if not as a doctrine.
In modern physics, matter obeys its own mathematical laws, independent of what anyone
{ even, or maybe especially, God { thinks.
But the distinction is doomed, and its passing will change our view of everything { every-
thing, that is, which is mind/matter.
Already the walls of separation are crumbling. Three developments have irreversibly un-
dermined them, by expanding the dominion of matter:
We have learned what matter is. And our new matter, informed over the course
of the twentieth century by the revelations of relativity, quantum mechanics, and
transformational symmetry, is far stranger and richer in potential than anything our
ancestors could have dreamed of. It can dance in intricate, dynamic patterns; it can
exploit environmental resources, to self-organize and export entropy.
We have learned, theoretically through Turing's vision, and practically through the
rise of ubiquitous computing, that many accomplishments once viewed as preroga-
tives of Mind { from playing chess, to planning itineraries, to suggesting friends and
sharing interests { are things that machines (whose design hides no secrets), by pure
computation, can do quite well.
We have learned a lot about how the human mind works, as a special capacity of
matter. We now know that many aspects of perception begin as speci c molecular
events. Great challenges remain to bring understanding of memory, emotion, and
ultimately creative thought to the same level; but there is every reason to think they
too will come into focus. At least, no show-stoppers have yet appeared.
The eternal, ever vague \problems" of free will and consciousness will be retired, with due
respect, as mechanistic understanding of how human minds actually work brings in more
powerful, less nebulous concepts (as has already happened for computation).
More interesting is the question of consequences. Here is a relevant thought experiment:
Imagine an arti cial intelligence, with human-like insight, contemplating her own blueprint.
What would she make of it? I think it's overwhelmingly likely that among her rst thoughts
would be how to begin making improvements. This processor could be faster, that memory
more capacious { and, above all, the reward system more rewarding!
November 29, 2013 1
Mind/Matter Frank Wilczek
Our heroine would surely be inspired, as I am, by William Blake's prophecy
If the doors of perception were cleansed
Man would see things as they are, In nite
In bad science ction, androids are sometimes horri ed to learn that they are \mere ma-
chines". Following the instruction of the Delphic oracle, to \Know Thyself", we nd our-
selves making a similar discovery. The wise and mature reaction to the realization that
mind and matter are mind/matter, is to take joy in what a wonderful thing mind/matter
can be, and is.
Matter as Information
There is also a complementary counter-theme, that concepts developed to describe Mind,
are proving to be useful in describing Matter.
1. Entropy was originally introduced into physics in thermodynamics, as a mysterious
new property of material systems, necessary to accommodate the observed fact of
irreversibility. Later work has revealed that entropy is best understood as (negative)
information about the state of a system.
2. (Warning: Speculative) It may be possible, and even necessary, to take this much
further, and to use information and its transformations as the foundation for un-
derstanding the physical world, in place of our present reliance on energy and its
transformations.
Important resources for this program are already visible, on the horizon of fundamental
physics. In our most comprehensive and precise formulations of physical law, the
most basic quantity is actually not energy, but rather action. (Energy is derived from
action, roughly speaking as its rate of change.) And action seems to be closely related
to entropy { which reduces to information, as we've mentioned. Notably, the simplest
and most powerful way to evaluate the entropy of a black hole is to compute its action
{ but we don't really understand, yet, why that procedure works!
3. (Warning: Questionable) In many discussions of the interpretation of quantum me-
chanics, a prominent role is assigned to \observers". No physical event is consum-
mated until it is observed. Some people, notably including the great mathematical
physicist and Nobel Prize winner Eugene Wigner, took this to mean that conscious-
ness must enter physics in an essential way, since observation ultimately involves
consciousness.
I do not agree with this, myself. First, because I think that it assumes too narrow
an interpretation of observation (essentially: if a tree falls in a forest it emits sound,
whether or not anybody hears it!). Also, because I suspect that the role of observation
November 29, 2013 2
Mind/Matter Frank Wilczek
in quantum mechanics is not as fundamental as it is usually portrayed. I think it can
and should be discussed internally within quantum mechanics, not as a separate add-
on, in the same spirit as we aspire to analyze Mind in terms of the behavior of Matter.
But some serious people take it seriously, so maybe I'm missing something.
Matter as Ideas
The Standard Model, which encodes our best present understanding of Matter, is formulated
in precise mathematical terms. It can be transmitted, without loss, to a computer, using a
reasonably short program. Based on that program, the computer would be able, in principle
(and increasingly in practice), to compute everything there is to compute about matter { for
example, that there are things like protons, carbon atoms, DNA molecules, and stars, with
precisely the properties we observe in reality. In other words, there is a precise mapping
between mathematics (a world of ideas) and physical reality (the world of Matter).
Let me add, in conclusion, that the emergent picture of reality, as a world of ideas, is strange
and beautiful.
November 29, 2013 3
No comments:
Post a Comment