注册 登录  
 加关注
   显示下一条  |  关闭
温馨提示!由于新浪微博认证机制调整,您的新浪微博帐号绑定已过期,请重新绑定!立即重新绑定新浪微博》  |  关闭

英海听涛

启用你的听觉,更自然地感受更辽阔的世界。——用进废退!

 
 
 

日志

 
 
关于我

喜欢科学,偏爱自然科学。也喜欢技术,儿时喜围观打锡壶的和锔盆锔锅的。当过工人,当过兵,第二故乡是湖南、广西和广东。

网易考拉推荐

欧洲的旗舰科学工程之一  

2013-02-13 11:17:17|  分类: 生命与健康 |  标签: |举报 |字号 订阅

  下载LOFTER 我的照片书  |
 

(1)

将来,用计算机完全模拟人脑是可能的。听起来让人难以置信,但是本周欧洲(共同体)委员会正式宣布选择人类脑工程作为其旗舰科学工程之一。

“一千亿个神经元,由一百万亿个突触所连接,人脑是我们所知道的最复杂的机器。欧洲研究者提出了一个大胆的研究人脑的新方法,他们的想法不复杂:用一台超级计算机来模拟一个完全的人脑。”

这一工程计划用时10年,花费超过15亿美元。用这些钱,我们许诺得到一个全新的方式来研究脑子,疾病和药物是如何作用的,以及这个神秘的器官是如何工作的。

——《科学在行动》,BBC 2013/1/31

(原话听写)

(06m14s)

In the future, it would be possible to simulate a human brain entirely in a computer. It sounds unbelievable but this week the European Commission has officially announced the selection of the “Human Brain Project” as one of its flagship science projects.

“A hundred billion neurons connected by a hundred thousand billion synapses, the human brain is the most complex machine we know of. European researchers propose a radically new approach to study the human mind. Their idea is simple: to simulate a complete human brain in a supercomputer.”

And the project plans to last ten years and cost over one and a half billion dollars. For that money we promise a whole new way to study the brain, how diseases and drugs affected and this mysterious organ works. –Science in Action, BBC 2013/1/31

(2)

“人类脑工程将聚集我们现有的全部知识,以便我们能一点一点地重建人脑。”

“这有点像摆弄一个有一万亿块碎块的拼图,而且大部分的碎块丢失了。我们打算要做的是在连接??的地方实行一种新策略,实际上以推测的方式填充拼图的空缺部分。我们要用我们今天所具有的数据来建立它,因此只是一部分数据。我们还要应用生物学的规则,一些规则告诉我们有多少神经元,或有多少连接,或有多少蛋白质进入一个细胞……我们要使用所有这些规则并且要创建一种算法,用这种算法把所有的数据纳入考虑。必须找到一种解决方案,生成一种单个神经元的模型,并最终建立整个脑的模型。它将与我们在过去200多年来所积累的所有数据和知识完全符合。

(07m24s)

“So the Human Brain Project will pull together all the existing knowledge that we have, so that we can reconstruct the brain piece by piece.”

“It's a bit like trying to build a puzzle with a trillion pieces, but most of the pieces are missing. What we are proposing to do is to begin a new strategy where we connect ???, filling of the pieces of the puzzle by actually predicting them. We try to build it with the data that we have today, so it's partial data. And we use biological rules, rules that tell us how many neurons there are or how many connections there are, or how may proteins go into a cell. And we use all these rules and we build an algorism, and the algorism then takes all the data into consideration. And it has to find a solution where it produces a model of a neuron and eventually a model of the whole brain. That is entirely consistent with all the data and knowledge that we have accumulated over the past two hundred years.”

(3)

“人类脑工程是要获得脑的细节,如同在生物学中观察到的尽可能精确地重建它。”

“但是构成脑的微小成分那么多,你曾描写过一些,所有这些因素都进来,能有足够的计算能力来模拟每一个这些东西吗?”

“在未来10年或甚至20年都不会有足够的计算能力来模拟每一个分子的相互作用。但是我们有应对策略,一旦你有了一个神经元的、或者神经元之间如何通过突触通讯的详细资料,那么它是可以抽象概括的。这样,你可以有部分模型工作在极其细节化的状态,而部分模型工作在低分辨率水平上。”

(08m55s)

“The Human Brain Project is to capture the detail of the brain, to reconstruct it as accurate as you can observe in biology.”

“But there are so many tiny parts that make up the brain, some of which you described, all these factors that come in, will there ever be computer power enough to be able to model every single one of those things?”

“In the next ten years or even twenty years there will not be enough computing power to model every molecular interaction. But we have strategies, once you have the detailed picture of one neuron or how neurons communicate through synapses, then this can be abstracted. And so you can have parts of the model running in great detail and parts of the (model running?) at much lower level of resolution.”

 

(4)

人类脑工程实际上由三个部分组成,我们刚才所说的其实只是其中的一个部分,可以被称为未来神经科学。

这第二部分称为未来医药。在这一部分我们的目标和任务是收集尽可能多的关于不同疾病的数据——从基因水平、分子水平、脑结构水平和血液水平。然后我们按照一定的规则数学地比较这些不同的疾病,力图找到各种疾病的特征。拥有了这种模型,我们就能探索并寻找各种事件的因果关系链,疾病是如何出现的?我们可以试验虚拟药物,寻找治疗疾病的最佳途径。我们还将有一个在今天还达不到的细节水平上模拟脑病的新途径。

还有我们称为未来计算技术的第三分支。我们能利用非常细节化的模型而且我们能简化它们,我们可以采用脑的“电路”。比如说你用的是脑的一个特定的被称为海马的部分,这个部分就如同谷歌的搜索引擎。这样,如果你拿来海马的“电路”,再简化它,你把它放到类似计算机芯片的硅片上。这“电路”是编了“程序”的,这芯片工作起来如同脑的一小部分。它不是整个的脑,它只是脑的一部分。这芯片就变成了一个搜索引擎。这就是我们所谓的神经形态计算技术。这样我们就有了类似人脑的能力,它们能对非常复杂的问题作出快速决定。你不必为想让它干什么而编程——它会学习。

——《科学在行动》,BBC 2013/1/31

(09m58s)

The Human Brain Project is actually composed of three parts. What we are talking about is actually only the one part. It could be called Future Neuroscience.

Then the second part is called Future Medicine. And in this part what we are aiming and doing is collecting data about as many different diseases as possible, so from the genetic level, and the molecular level, and the brain structure level, and blood level. And then we compare between these different diseases mathematically and based on rules. And we try to look for signatures of diseases. With such models we can now explore, then we can look for the causal chain of events will do, how the diseases came about? We can try out virtual drugs to see what’s the best way to fix that disease. We'll also have a new way to simulate brain diseases at level of detail that is just not being possible today.

And then there is a third branch, which is what we call Future Computing. A way we can take these very detailed models and we can simplify them. We can take circuits of the brain. Let's say you take a certain part of the brain that's called the hippocampus, this part is like a Google search engine. So, if you take the circuitry of the hippocampus, and you simplify it, you put it onto a silicon chip like a computer ship. The circuit is programmed and the chip works like a little piece of the brain. It's not the whole brain; it's a piece of the brain. And then does the chip become a search engine. And this is what we call neuromorphic computing technologies. And then we would have brain-like capabilities. So they make very fast decisions on very complex problems. You don't program what you want it to do, it learns.

--Science in Action, BBC 2013/1/31

  评论这张
 
阅读(143)| 评论(1)
推荐 转载

历史上的今天

在LOFTER的更多文章

评论

<#--最新日志,群博日志--> <#--推荐日志--> <#--引用记录--> <#--博主推荐--> <#--随机阅读--> <#--首页推荐--> <#--历史上的今天--> <#--被推荐日志--> <#--上一篇,下一篇--> <#-- 热度 --> <#-- 网易新闻广告 --> <#--右边模块结构--> <#--评论模块结构--> <#--引用模块结构--> <#--博主发起的投票-->
 
 
 
 
 
 
 
 
 
 
 
 
 
 

页脚

网易公司版权所有 ©1997-2017