主题：【原创】围绕脑科学而发生的若干玄想 -- 鸿乾
分两个部分：一个部分是如何把某些智能的重要方面机械化，例如学习成了机械学习，或者机械辅助的学习，一个部分是动机，更重要的复制。前面的那个部分，是完全可能实现的。但是，后一个部分是不可能实现的。简单一句话，这个部分是若干亿年的进化的结果，任何所谓奇点或者机器，是不能在这一部分有所作为的。这位作者，不是深入研究前一部分，大概他也不具备研究前一部分的能力，就夸张后一部分来争取卖点。
推荐大家看两个东西。
1. 央视的成语大会的决赛：外链出处 大家可以看到联想，词语的变形，等等。这是很高的智力吧？
2. 澳洲一家公司的网页：外链出处 大家可以看到自然语言的运算，例如：Apple 减去 Fruit 等于 Computer，也就是说联想，词语变形等高智力的东西，其实是可以机械计算的。就是说，完全可能几十年后，机器可以也来玩这个成语大会的游戏，而且远远超过人了。
题外话，游戏中，中国男孩远差于中国女孩，就是说这类智力游戏中，男子完全不如女子。但是在现实生活中，是什么情况呢？这也是一个对我最前面的观点的支持。
已经成建制.
1.
human brain is "too hot"
actually, Penrose talked about the disadvantage of human brain as a quantum computer model: it often goes too hot.
Pensose's some books are all translated into Chinese. He has ground breaking work in 准晶体 area.
准晶体- 维基百科，自由的百科全书
zh.wikipedia.org/zh-hk/准晶体
轉為繁體網頁
但是直到快20年后这种理论上的结构才和准晶的研究联系起来。自然界中非周期 ... 一种典型的準晶体结构是三维空间的彭罗斯拼图（Penrose）。二维空间的彭罗斯
and I talked about that too.
"熱輻射"對大腦的"傷害" [ 晓兵
2.
in many 具体行业, "self-aware" robots are already replacing human beings as "employees".
it is not only about "cost": "self-aware" robots do not have"大腦 too 熱" issue, and can work a lot of better.
减法 operation, a brand new challenge to global capitalism.
3.
fundamentally, to make "2045年最聪明的种族不再会是人类":
that is a challenge to quantum physics, at 分子 first, many bio 分子 are not even 准晶体, very hard for physics.
in that sense, 基因食物 does not have real physics support, still a largely 唯象理论;
唯象理论 often are more profitable, 水至清则无鱼, politics is the worst 唯象理论, but is most profitable in Tgchina, but that is a 内战内行 外战外行 model, not going to work internationally.
US 包圍 tgchina: more of 物理学 now
------------
http://www.louisdelmonte.com/will-strong-artificially-intelligent-machines-become-self-conscious-part-22-conclusion/
Part 1 of this post ended with an important question: “How can we determine whether an intelligent machine has become conscious (self-aware)?” We will address this question in this post, along with some ethical dilemmas.
We do not have a way yet to determine whether even another human is self-aware. I only know that I am self-aware. I assume that since we share the same physiology, including similar human brains, you are probably self-aware as well. However, even if we discuss various topics, and I conclude that your intelligence is equal to mine, I still cannot prove you are self-aware. Only you know whether you are self-aware.
The problem becomes even more difficult when dealing with an intelligent machine. The gold standard for an intelligent machine’s being equal to the human mind is the Turing test, which I discuss in chapter 5 of my book, The Artificial Intelligence Revolution. (If you are not familiar with the Turing test, a simple Google search will provide numerous sources to learn about it.) As of today no intelligent machine can pass the Turing test unless its interactions are restricted to a specific topic, such as chess. However, even if an intelligent machine does pass the Turing test and exhibits strong AI, how can we be sure it is self-aware? Intelligence may be a necessary condition for self-awareness, but it may not be sufficient. The machine may be able to emulate consciousness to the point that we conclude it must be self-aware, but that does not equal proof.
Even though other tests, such as the ConsScale test, have been proposed to determine machine consciousness, we still come up short. The ConsScale test evaluates the presence of features inspired by biological systems, such as social behavior. It also measures the cognitive development of an intelligent machine. This is based on the assumption that intelligence and consciousness are strongly related. The community of AI researchers, however, does not universally accept the ConsScale test as proof of consciousness. In the final analysis, I believe most AI researchers agree on only two points:
1.There is no widely accepted empirical definition of consciousness (self-awareness).
2.A test to determine the presence of consciousness (self-awareness) may be impossible, even if the subject being tested is a human being.
The above two points, however, do not rule out the possibility of intelligent machines becoming conscious and self-aware. They merely make the point that it will be extremely difficult to prove consciousness and self-awareness.
Ray Kurzweil predicts that by 2029 reverse engineering of the human brain will be completed, and nonbiological intelligence will combine the subtlety and pattern-recognition strength of human intelligence with the speed, memory, and knowledge sharing of machine intelligence (The Age of Spiritual Machines, 1999). I interpret this to mean that all aspects of the human brain will be replicated in an intelligent machine, including artificial consciousness. At this point intelligent machines either will become self-aware or emulate self-awareness to the point that they are indistinguishable from their human counterparts.
Self-aware intelligent machines being equivalent to human minds presents humankind with two serious ethical dilemmas.
1.Should self-aware machines be considered a new life-form?
2.Should self-aware machines have “machine rights” similar to human rights?
Since a self-aware intelligent machine that is equivalent to a human mind is still a theoretical subject, the ethics addressing the above two questions have not been discussed or developed to any great extent. Kurzweil, however, predicts that self-aware intelligent machines on par with or exceeding the human mind eventually will obtain legal rights by the end of the twenty-first century. Perhaps, he is correct, but I think we need to be extremely careful regarding what legal rights self-aware intelligent machines are granted. If they are given rights on par with humans, we may have situation where the machines become the dominant species on this planet and pose a potential threat to humankind. More about this in upcoming posts.
Do you have to always relate this to something irrelavent, such as politics?
晓兵 is a robot!!
Obviously, 晓兵 had denied that.
1.
I care about TG the most, because TG is potentially the biggest game changer to this US led game;
if all potential game changers are neutralized, then the systematic risk is reduced significantly;
US led 2008 global financial crisis: part of reasons, globalization of capitalism, china's rise and its huge impact on the system, etc.
now the system is somehow adjusted itself to the huge disruption caused by Tgchina as a huge new elephant in the otherwise already crowded house, and this USA managed house has largely stabilized, although at one point it almost falls apart in 2008.
that is the #1 reason behind the current US bull market: tgchina's disruption to the system has been digested fairly well overall, and as a result, largely priced in by financial market.
kind of why Tgchina's economic mess does not really bother global capital market, the way EU did.
the global capital market kind of has come to a fairly good understanding of Tgchina as an significant player in today's world.
2.
but now we have a new evil, this "self-aware" robot army, partially because of that, low "labor participation rate" and "low wage growth rate" in US, still bothering US Fed big times.
then what the heck that has to do with tgchina?
it is all relative in today's global market: if Tgchina suffers more, USA will suffer less, the so called the "lessor of the two evils", then this on-going US bull market will be ok, in the sense that there is not so much systematic risks going on, vs 2008-2010
then you buy the big dip when it comes, with little risk of getting a falling knife coming onto your head.
in that sense, "all ALGOs have to make $" in almost everything I comment about, or I would not care about it.
3.
各取所需
because of my native language is Chinese, I found it extremely helpful with some posts, many of them I have quoted many times, helpful in the following sense:
for example, I am supposed to be a fiancé guy, but a few posts here in this forum articulated finance theory's ALGOs much better than I could, and similarly with many other social science ALGOs;
as said, "white" social science ALGOs are hard, if not articulated well and it is in English, it is even harder.
but as said in "刘亚洲:"改變國際舊秩序的,不是蘇聯.而是美國", 刘 talked so well in our mainlander's language, concept and model;
similarly, many posts here really get me "clicked" on some white's religion and other social sciences.
4.
and why do I bother myself with those things?
white often thinks with same ALGO, across many areas, physics, finance, religion, etc.
if I want to model white's financial market, I have to learn about those other white's ALGO as mentioned in the above;
5.
then why do I care about all these Chinese stuff?
other than the reasons as said in the above, comparative modeling of white vs Chinese ALGO really helps me in understanding human mind ALGO in general, then it would be more physics ALGO like, (physics laws are independent of the coordinate system one picks)
I wish I could type fast enough in Chinese and write "less strange", then I would get a lot of feed back and discussions, then I would have come across even more "mind-blowing" posts, so I can understand "white evils" much better, particularly the white's finance game, the only thing I really care about, in a way.
before that, there would be almost no readership of my strange posts in English, with a lot of physics jargons.
6.
still, I will say it again and in a general sense: have your kids start on physics and English, regardless of what they do when grown up.
social science is getting "normalized" by physics ALGO, piece by piece, but it is getting there, and as a result, social science jobs will be becoming more and more physics jobs, kind of.
basically, it is the same ALGO runninbg: if social science BS cannot help system make $, it will be hit by system's 减法 operation, one way or another. kind of why I don't bother with me with most posts here in this forum, except for a very few areas.
kind of another "weak area" of Tgchina's system, vs USA, who imports a lot of physics brains from abroad, while more and more white kids become white trash.
7.
an example:
white's religion ALGO
1511, 領悟“神的義”與“人的信”, 馬丁·路德: it is a paradigm change or ALGO change if you will, and
"诸神的黄昏": 资本把人类卷入社会化大生产 , my that post quoted and it was totally inspired by the following one;
http://www.ccthere.com/article/4018881
如果跳出中国，远离政府看历史的话 [ 喝点红茶上会网 ] 于:2014-06-08 21:41:26 复:4018800
that new white religion ALGO basically neutralized religion as a "social science disruptor" to the emerging industrial capitalism in Europe, a new 生产关系 for white's world;
It tells all kind of religion 政治局常委: get out of way of industrial capitalism, do whatever the heck of god stuff in your private house, but never attempt to make it to the state level or social level: we are busy making machines, cars, etc, not the stupid 聖經.
That 馬丁·路德 new religion ALGO was obviously a game changer, although it may not what he intended to do, it does not matter:
system ALGO will do whatever it needs to do, killing or rewarding any individuals who happen to be "at the wrong place and wrong time", or those super lucky ones who got hit by money falling from the sky.
the super power of system's ALGO.
8.
and now, industrial capitalism as a 生产关系 is being replaced by informational and/or ALGO capitalism as articulated by
汉密尔顿ABC"講金融, "humanity 自身的知识和勇气" 的稀有性 (value) and pricing, a core ALGO of today's financial capitalism.
9.
now, Tgchina's #1 super ALGO, "毛林共识", its fundamental problem is: it is nothing but a huge collection of "outdated software": Chinese traditional ideology such as 刑不上大夫, Marxist "economics", and 列寧 "國家與革命", and perhaps, some chapters from WTO.
knowing that super ALGO/TG 聖經 is already outdated, TG now tries to package it up with china dream, 中国文明 and other just last night invented whatever social science BS.
10.
at system level, I think TG knows about all these issues, and so does global capital market, and those super smart traders.
how TGchina is going to evolve from here on? it could be anybody's guess, but as said, Tgchina's disruption to the uncle sam managed global capitalism HOUSE has largely topped out, with Tgchina as a systematic risk factor.
and as said, for whatever reasons, tgchina is not likely to fall apart overnight, causing a global earth quake economically or financially.
in that sense, whatever BS noise coming out of Tgchina, are not as loud as it used to be:
BS=BS, period.
TG knows that: Li went to London to bribe British.
TG's asset "premium": out, now what discount you are going to use to price TG BS?
11.
obviously, to TGchina, those propaganda are part of their super "毛林共识" ALGO run:
"熱輻射"對大腦的"傷害" 晓兵
TG has to do that, just like TG has to build GFW.
and as an individual, you do not want to be at "wrong place and wrong time" when TG's super ALGO is running wild, TG has to do whatever it has to do now, or Tgchina may have fallen apart yesterday.
and again, have your kids started on physics and English, making better sense now?
12.
"self aware" and self correction: it is not about AI, it is about almost any systems.
as said, "毛林共识" had worked out for many years for TG, until recently, that is why it is TG's super ALGO, everybody has to run it, it is in TG's gene.
13.
此一時彼一時, now "毛林共识" is outdated already, rotten out from inside.
where is new software? everybody is running new software, where is your?
14.
the challenge.
Is Tg "self-aware"? may be, at least some TG senior traders know what is going on;
can TG "self-correct"?
does TG have a system to mobilize the brain power of huge size at various level of Chinese society to figure out the path to "self-correct" etc?
not really, if your "毛林共识" ALGOs keep putting some of those brains into jail, only because they dare to stick their necks out and speak out a different ALGO other than "毛林共识".
and with GFW: even your smart brains are very likely day dreaming, anyway.
15.
TG as a system: "self aware", may be; "self-correction", very unlikely;
TG system with "毛林共识" as its core ALGO =落后的生产关系, relatively to Uncle Sam's ALGO, period.
and 落后=to be fxxed, one way or another, sooner or later.
now, how do you place your trades?
16.
why do I care about this AI thing?
there is ALGO, but no 没有阴谋论 in social system, fundamentally.
as said, physics cannot even model a 基因 as a system of millions atoms working together, how anybody can possibly have a consistent 阴谋论 model for a social system?
it is just impossible.
because of all these, 生产关系 business is hugely profitable business, with a huge potential for arbitrage.
and these AI and brain stuff help us to model human mind ALGO, particularly how human minds tend to work together.
and human minds have been working together for millions years already, and that ALGO is still running, and it does not change much at all.
相对论方程"一阶性".
OMG, the great 愛因斯坦, although 洛侖茲和龐加萊 etc had put together some major pieces already for 狹義相對論.
but, system's ALGO can't remember too many names, and 愛因斯坦 got picked for SR, GR and many many physics ALGOs. He is GOD. period.
The system's ALGO is too busy with money making, 勞動生產率 etc, and everybody tries to guess that super ALGO's next step, then you ride it, making good money, and have fun.
本帖一共被 1 帖 引用 (帖内工具实现)
1.
self-aware: an important concept, thanks, and I read his blog a little bit.
penrose and Louis Del Monte, and others all talked about "self-aware" aspect of 人类 "聪明" vs AI.
in analogy, "self-aware" and "self-correction" of 人类 "聪明" vs AI Robot is like 高阶 logic system vs 低阶logic system;
although there are many 高阶 logic system in terms of math, fundamentally "a 高阶 logic system" is a physics challenge:
can an human made physics system know how to survive a 高阶 disturbance, and where is the border of that 高阶 disturbance, in terms of system's survival?
or in terms of 相变, can a physics system know which 相 to jump into, if it has to 革自己的命 (almost like TGchina, relevant(:)?)
obviously, we are getting into quantum physics, etc.
quantum physics "bottle neck": 纳米材料 etc breakthrough, very hard; otherwise, 宏观量子效应, quantum computer etc=theoretical concepts;
before that "break through", now AI is doing "self-aware" "self correction" in "math" software, basically monte carlo simulation/big data, still a huge potential: we barely started.
2.
so far, I have not seen any areas where it does not make sense at all, in terms of borrowing some physics ALGO to help figuring out the game;
those borrowed physics ALGOs all run "under the hook", much like a software engineer does not have to learn quantum physics to understand how a semi conductor becomes a computer hardware
我今天看到河里的这个帖子：
感到有些意思，大家也都可以看看。说几点：
1. 这个虽然是商业动机的视频，大家还是可以看看，了解一下。
2. 是否有对左脑和右脑有研究的朋友？请给我们多做一些比较深入的介绍。
3. 这个生意我自己感到是以后的很多生意的雏形。运用最新的脑科学，来编成若干可以影响脑发育的教程，进而影响各种学习的教程，是一个方向，有很多潜力。这个里面肯定有很多花头和虚浮的东西，而且也是一个逐渐实验的长期过程。但是，肯定有人搞，也值得搞。
4. 那个让小孩猜测图片的游戏，我想基本上就是训练小孩察言观色的能力的游戏。
5. 那个小孩可以记住很多词语，简直就是把人脑当作数据库来做，也反映了很多人脑记忆的特征。是否这个东西对小孩的智力有帮助，他们说有，我不知道，也不相信，但是也没有理由否定。但是，这个事情，的确反映了人脑记忆的很多特征。
6. 其实，我想，这样的教程，也不一定仅用于认知能力这样的智力开发，也可以用于体能，技能等的开发，例如运动得很好，打球很好，滑雪很好等等。
Secrets of the Creative Brain: 外链出处
还有PBS的视频： 外链出处
这是最近的大西洋杂志的文章。颇长。值得读读。
简单评论几点：
1. 那个图，就是创造性大脑和平常大脑的比较，令人震撼。现在的确有人找到了“具有创造性”的物理证据。注意，是物理证据。就是说，具备创造性，其实就是大脑硬件/软件有所特殊，使得这些大脑的图样识别器更敏锐，更发达。我想，这个认识，对于教育学一定会产生重大影响。
2. 看这段话：
3. 不过，我相信，当脑科学发达后，特别是机械学习机器大量运用的时候，人其实可以把那些图样识别器交给机器来做。
4. 具备自学能力和自学经验，是极端重要的。我想，这个认识也将对教育学产生重大影响。
5. 文章中的评论：
本帖一共被 1 帖 引用 (帖内工具实现)
IBM's TrueNorth processor mimics the human brain 外链出处
去年我在这里给过他们的这个芯片的第一代的视频：链接出处
极低的耗能是这个芯片的关键。我猜想，他们的这个芯片的时钟频率不高，可能在每秒几百。但是，因为是类脑计算，每一个时钟频率，可以做很多“识别”上的事情，这些事情，如果让现有的软硬件来做，就需要上百亿的芯片周期来做，因此极端耗能。
如果这个东西成型，现有的计算模式很多都需要改变。
把这样的芯片用到无人机上面，的确是非常合适的应用。两个要求都使得这一的应用是其他的软硬件不如的：1，低功耗要求，2，对使用环境有很多未知，因此导致控制模式的很多未知，使得编程不容易。这个芯片和下面报道的IBM的芯片，其实有很多共同之处。
该神经形态芯片具有576个神经元，能够对由光学、超声和红外传感器采集的数据进行分析和做出响应，该芯片能在测试房间之间飞行时对本身进行自动重新规划。
麻省理工技术评论指出，重量为100克的无人机在3个不同的房间飞行，并验证“从墙壁、家具和其他物体传入的传感器信号的独特模式，可引起神经元里的电子活动模式，而该芯片此前从未有过类似经验。”该芯片能够学习房间布局，并在无人机下次访问某个房间时做出适当的反应。
休斯研究实验室神经和新兴系统的主任Narayan Srinivasa表示：“这表明，即使面临着严格的尺寸、重量和功率限制，在飞行过程中进行学习也是可能的。”他同时表示，通用动力和波音公司正在考虑将这项技术商业化。
一家位于硅谷的Aerovironment公司，已经为国防部构建了6平方英寸的原型无人机，包括电池在内重量只有93克。该芯片本身重量仅18克，功耗50毫瓦。
该新型无人机和芯片通过DARPA的“神经形态自适应可塑可扩展电子系统”(SyNAPSE)项目开发，该项目的目标是实现模仿生物大脑活动的技术，并应用于机器人。在神经系统中，突触是一种结构，能够允许神经元传递电子或化学信号到另外的神经元。
SyNAPSE项目由休斯研究实验室、惠普公司和IBM公司合作开展。该项目始于2008年，IBM公司从DARPA获得了4200万美元资助，而休斯研究实验室被授予3450万美元。（工业和信息化部电子科学技术情报研究所 王巍）
逻辑人生--王浩小传 链接出处
以前没有看到这个文章。感到很好，推荐给还没有读过的河友。王浩是老一辈留学生中的佼佼者。如果有人来写王浩的传记，也是有意思的。他的关于图灵机的工作，现在仍然非常有意义。
数理逻辑是玄之又玄的学问。但是，恐怕大多数人都忘记了，或者根本就不知道，数理逻辑其实是现在的最炫目的时尚玩意的最重要的思想源头，而且还将继续是这些时尚玩意的思想驱动。就是说，如果没有那些完全不食人间烟火的纯学术研究，今天的炫目时尚也就没有了源头和驱动。如果没有莱布尼兹，康托尔，弗雷德等等开创的数理逻辑，没有希尔伯特的那些问题，很可能就没有图灵和丘奇的工作，因此也就没有了计算机的基本源头之一。那么现代是否还有计算机呢？肯定有某种形式的。IBM的卡片机，就是在图灵之前。但是，那不是通用计算机。因此很可能，如果没有图灵和丘奇的工作，没有冯诺曼的工作，就只有某些非通用的计算机，现有的一切高科技，都完全无法设想。
这里面的教益非常丰厚。如果有人愿意来写王浩的传记，那就是从中国人的角度出发来总结这里面的教益，肯定很有趣。
图灵机非常成功，但是，我认为，现在的历史已经到了跨越图灵机的时候了。图灵机是通用计算机，就是说，一切可以计算的东西，都可以通过这个机器来计算。这样看来，好像是无法超越的。的确这是无法超越的。不过，跨越不是在这个方面，而是在机器学习。我认为，我们正站在这个新时代的门槛前。究竟谁先突破这个门槛？
河里有很多好帖。但是都无法检索。仅有某些时候，自己冒出头来，我们才能看到。这是一个西西河的重大弱点。
本帖一共被 1 帖 引用 (帖内工具实现)
前些天读到这样的一些东西，举例来说，这个链接：外链出处
感到有趣，评论一下。先看这段话：
Thatcher’s study suggests a counterintuitive notion: the more disorganized your brain is, the smarter you are.
就是说，脑内的若干区域，有的时候是所谓的phase-lock，就是说脑细胞按照一个韵律活动，就像芯片中的那样，但是，同一区域，有的时候是混乱的，脑细胞就没有按照韵律活动，而是各自活动。这两种方式交替出现。据测试，平均的混乱活动时间是55毫秒。有趣的是，这个混乱活动时间在个体之间有明显变化，但是更加有趣的是，这个混乱活动时间的长度，直接和智商相关。混乱时间每多一个毫秒，智商就高20个点。
这是非常惊人的发现。我不知道这个发现在脑科学专业里面是否得到了全面的验证，以及基本的赞同。如果的确这个发现得到了认同，那么这将是对人脑的认知能力的很重要的发现。这就是说，人的认知能力，直接和脑内的生理活动的具体结构联系了起来。
原始的论文在这里：外链出处 不知是否有专业人士可以来给我们讲讲更多，更具体的有趣事情。
那么，这样的事情，大概也将在机器学习中反映出来。当然，我并不知道具体如何反映。
先看报道：外链出处
openworm这个计划已经做了很久了。但是这次装上机器虫，还是第一次。值得继续关注后面的发展。
不管怎么说，这次把软件装上了机器虫，这个机器虫的行为就不是通过编程来达到了，而是通过机器虫和环境的互动，来使得机器虫学习环境，其行为就是通过学习来达到。就是说这个机器虫是一个有效的学习机器。
这个虫仅有302个神经元，是最小的神经元连接体（connectome）。
按照排列组合有：既学又思（11），学而不思（10），不学而思（01），不学不思（00）。
学是指读书看报看帖等，是从外部向大脑输入信息；思是大脑内部处理信息。输出信息对应说写回帖等。
我很喜欢你的这个解释：学是指外部输入信息，思是内部处理信息。那么一个信息处理盒子（我们就暂时不用机器这个词，因为人脑不是机器，但是也是信息处理盒子），必然的状况就是学思四合之一。
我认为，除开不学不思是我们不需要的状况外（其实也需要，都需要休息，就是机器也需要休息），学而思，学而不思，不学而思，其实都是需要的，都有其独特的作用。我认为，在认识论上都有重要的作用。
谢谢你的这个总结，学思四合，好。