何毓琦的个人博客分享 http://blog.sciencenet.cn/u/何毓琦 哈佛(1961-2001) 清华(2001-date)

博文

The “Next Big thing” - Must Read 下一个大发明– 必读 (中英文) 精选

已有 20450 次阅读 2009-4-29 11:12 |系统分类:海外观察

(Note added 4/28/2010 There is a 20 minute video by Steve Wolfram talking about a computational theory of everything using Mathematica and Wolfram/Alpha  http://www.ted.com/talks/stephen_wolfram_computing_a_theory_of_everything.html?utm_source=newsletter_weekly_2010-04-27&utm_campaign=newsletter_weekly&utm_medium=email 
(Note added 11/14/09 Popular Science, the world's
largest science and technology magazine, has released its 22nd annual list of the best 100 innovations, and named Wolfram|Alpha as the "Best of What's New" Grand Award winner in the category of computing. Popular Science states that all 100 innovations must
"push past what we thought was possible,"
visit the Wolfram|Alpha Blog to read the full announcement:

http://url.wolfram.com/1d.bMvO/
(Note added 6/25/09 A detail article explaining Wolfram/Alpha can be found in the latest issue of Technology Review, July/August 2009
(Note added 4/29/09 10:30am EST:
The Berkman Center for Internet & Society has posted a recording of
yesterday's webcast on its site at
http://cyber.law.harvard.edu/interactive/events/2009/04/wolfram)
Before you start reading this blog, I must ask you if you have ever heard of the person “Steve Wolfram”, or the book “A New Kind of Science”, or the software “Mathematica”. On the rare chance you have not heard/seen/used all three, please google one of these topics and read about it before your start.
In information technology, everyone is always looking for “the next big thing” or more narrowly or specifically, the next “killer application”. No one would disagree that WINDOWS and GOOGLE were the last two “big things”. I am going to tell you about “the next big thing”. On 4/28/09, EST 3:00pm Steve Wolfram gave a prelaunch presentation at Harvard University of his next project titled Wolfram/Alpha which he will announce to the world sometimes in early May 2009.
So what is Wolfram/Alpha?
Imagine you wish to know the detail weather on the day, April 19 1899 for the town of Lexington, MA, USA (my hometown actually) and the same for another town in Asia at the same latitude (ShenYang 沈阳 actually), and how they compare. Or you wanted to know something about the mathematical properties of the indefinite integral of x^2(sine of x)^3dx. By the way, the information basis of the answers to these questions actually exist in some data base available on the Internet. But can you search or google for it? A rather inexperience user of Google such as myself certainly don’t know how. Even a very advanced expert will require quite a bit time, several queries, and additional calculation and graphics after s/he obtains the basic data to answer the questions. Now imagine if you just type in
“weather Lexington, MA 4/19/1899” in the dialog box and the temperature by the hour, the weather and wind condition on that day were instantly displayed for you in graphical and tabular form.  And if you add “vs. a big town in China at same latitude” , the same information will displayed side-by-side in tables and superimposed as curve in graphical form. Similarly if you type in the mathematical form of the indefinite integral of x^2sinx^3dx, immediately a plot of the integral in graphical form and closed form analytical answers if any together with any salient property of this function appear as answers.
Additional examples questions that Wolfram/Alpha can answer and are demonstrated live in real time at the presentation are:
1.                 Type in “6000 C”
Answer: equivalent in Fahrenheit, what metal will not melt at such temperature, the temperature at the surface of the Sun, etc.
2.                 Type in “ LDL 180”
Answer: Distribution of Cholesterol level in the US population, What you need to do if this is your Cholesterol level, medicines to lower your cholesterol number, etc
If you add “age 40” to the dialog box then the answer further specialize to data for the “age 40” qualification and any other information, such as life expectancy, etc that Wolfram/Alpha thinks you may need. The point is that you can “DRILL DOWN” for more information. In fact the Wolfram/Alpha answer page will in addition to answers suggest various possible paths for you to ask further questions.
3.                 Type in any sequence such as “ATGTA. . . “
Answer: Wolfram/Alpha  will understand that this a genome sequence and will return whatever is known about this sequence – e.g., what is its place on the human genome, what biological function if known the sequence governs, etc.
4.                 Type in “CSC”
Answer: Wolfram/Alpha  understands this is a stock symbol on the NYSE for the stock of Computer Science Corp. Earnings, stock price, expert opinions for the past as well as projected future will be displayed.
5.                 Fish production of France vs. Poland
6.                 President of Brazil in 1982
7.                 Tide in New York City Harbor on 1/1/2015
8.                 Next total solar eclipse visible in Chicago, USA
Answer: 15 years from now, duration, and eclipse path plotted against a world map
9.                                           9.        What is the 500th largest country in the world?
10.                                      Answer: no such country or Wolfram/Alpha  does not know the answer to this query
These and many other queries were demonstrated live during the presentation. I think you will agree that GOOGLE cannot accomplish these answers without a lot of expert human help and only in non-real time. It other words “Wolfram/alpha promises to make everyone in the world an instant expert on anything”
So how does Wolfram/Alpha  do it?
For data it relies on the vast amount of information and databases that already exist on the Internet. For calculations and visualization, it utilizes the capability of MATHEMATICA. However, this is easily said than done. Wolfram Research, the company, employed over 100 persons for ten years to accomplish this project prior to unveiling it today (4/28/09) and publically launch it sometime in the next two weeks (Early May 2009).
There were four major components in Wolfram/Alpha :
(i)                Data curation – While vast number of database exist on the WWW. Most of them use incompatible format, different languages, and sometime with inconsistent and faulty data. Wolfram/Alpha  must first clean up, correlate, audit, and verify these DB and transform them into one uniform and consistent format before they can be access quickly. This is a time consuming and huge task
(ii)              Computational algorithms – This is the relatively easy part since MATHEMATICA is already well developed
(iii)            Language ability – While the AI problem of understanding general language remain unsolved, the problem of making sense of a query, even if it is ill formed, can be broken down into a finite set subproblems that can be tackled. We already see this is in “the automated telephone answering” software that is present in customer service popular with many manufacturer of equipments. In other words, we know when we initiate a query, we will not be making idle polite chit-chat with the computer or asking the computer if it feel sad/happy today. We only have a specific type of goal in mind. This makes the “understanding” free form language considerably easier.
(iv)            Automate presentation – This has to do with graphic user interface (GUI) and user friendly design. Again this aspect is well understood.
Putting these four tasks together and you have Wolfram/Alpha. The creator claims that as of today it covers 95% of the knowledge in a typical reference library. Of course, the project is on-going as more and more DB and capability are integrated into the system (Just as GOOGLE of the 90s are very different and far less capable than the GOOGLE of today).
Question and Answers from the audience.
How does Wolfram/Alpha deal with inconsistent, incomplete, and uncertain data? Answer: whenever possible, W/A provides original sources, warnings, footnotes, and ranges of uncertainty if applicable in results
Are there documentation of APIs for Wolfram/Alpha? Answer: there will be.  Of course, since Wolfram Research is a commercial company certain part of Wolfram/Alpha  will be proprietary.
Will you be able to personalize Wolfram/Alpha? Answer: yes, once you have access ot the APIs
What other information are provided on the Wolfram/Alpha  answer page? Answer: assumptions used in getting the answer and choices for further inquires
What is the business model for W/A? Initially it will be free. Later on we plan to have Ads (just like Google) and subscriptions for specialized users.
Will this presentation be available on the web? Yes in due time. (Please watch this blog.  I will post it as soon as I know. It is well worth 1.5 hours of your time. For more write up see http://www.readwriteweb.com/archives/wolframalpha_our_first_impressions.php
and google W/A) (Note added 4/29/09 10:30am EST:
The Berkman Center for Internet & Society has posted a recording of
yesterday's webcast on its site at
http://cyber.law.harvard.edu/interactive/events/2009/04/wolfram)
Two questions I pose to myself?
What makes “Harvard” Harvard? It is where important discovery and announcement are often made by the creator himself live before the rest of the world knows about it.
Why do I blog? So that you, the reader can say, I read about this first on Science Net.

这篇博文是关于我47年职业生涯中所听过/看过的最令人印象深刻的宣讲。(美国东部时间09429日上午1030添加注释:哈佛波克曼网络与社会研究中心已经在其网站上链接了昨日宣讲会的录音,见http://cyber.law.harvard.edu/interactive/events/2009/04/wolfram)

 

在你开始读这篇博文前,我必须要先问问,你是否听说过Steve Wolfram这个人?或者新科学?或者Mathematica软件?万一你还没听过/看过/用过他们,那么在往下读之前先用GOOGLE搜索一下吧。

 

在信息技术里,每个人总在寻找下一个大发明,或更窄一点更具体一点来说,就是下一个杀手级应用。没人会反对WINDOWSGOOGLE是最后的两个大发明。下面我要说的是关于下一个大发明。东部时间09428日下午3点,Steve Wolfram在哈佛大学就他的下一个计划作了上线前宣讲,该计划名为Wolfram/Alpha,他将在5月初对外宣布。

 

那么Wolfram/Alpha是什么呢?

设想一下,你很想知道1899419日美国列克星敦市(我的家乡)的详细天气情况,和同一纬度的亚洲一座城市(沈阳)那天的天气情况,以及它们之间的比较;或者你想知道x^2(x的正弦)^3dx的不定积分的数学属性。顺便提一下,这些问题的答案的信息基础实际上存在于因特网上的一些数据库里。你能搜寻或GOOGLE出答案吗?像我这样不太有经验的GOOGLE用户自然不知道怎么做。即使是专家,也需要一些时间和多次查询,在获得基本的数据后,还需要进行计算和画图来回答这些问题。

 

现在设想你往对话框里敲进“列克星敦,天气,麻省1899419日”,结果每小时的即时温度、当天的天气和风力情况立即排列成表展现在你眼前。而且如果你往里添加“对比同一纬度的中国的一个大城市”,相同的信息会并排显示在表格里,而且会叠加成曲线的形式。同样,如果你以数学公式的形式输入不定积分“x^2sinx^3dx”,很快就能得到绘图和封闭曲线形式的积分图标,如果有的话,连同这个函数的突出特性一起显示出来。

宣讲会上Wolfram/Alpha实时解答的另外一些问题例子包括:

1,敲进“6000C

回答:等同于华氏度,什么样的金属在该温度下不会熔化,太阳表面的温度等等。

2,敲进“LDL180

回答:美国人口的胆固醇水平分布,如果这是你的胆固醇水平,你该做些什么,降低胆固醇水平的药物等等。

如果你往对话框里添加“40岁”,那么回答就会更加具体化到符合“40岁”的数据,以及其他任何Wolfram/Alpha认为你可能需要的信息,比如寿命等。关键在于你可以“钻取”更多的信息。事实上,Wolfram/Alpha回答页面除了答案外,还会提供多种可能的途径以方便你继续问问题。

 

3,敲进任何一段序列,比如“ATGTA

回答:Wolfram/Alpha会识别出这是基因组序列,并显示出关于该序列的任何已知知识——比如,它在人类基因组中的位置,该序列的生物学功能(如果已知的话)等等。

4,敲进“CSC

回答:Wolfram/Alpha识别出这是计算机科学公司在纽约证券交易所的股票标志。将会显示利润、股票价格、过去专家观点以及预期前景等。

5,法国与波兰的渔业产量对比

61982年巴西的总统

7201511日纽约港的潮汐

8,芝加哥下一次肉眼可见的日全食

回答:15年以后,并在世界地图上标出持续时间和日食路径。

9,世界上第500大的国家是那个?

回答:不存在这样一个国家,或者Wolfram/Alpha不知道该查询的答案。

这些问题和其它许多查询在宣讲会上进行了现场演示。我想你肯定会同意,GOOGLE在没有很多人工帮助下无法得到这些答案,而且是非实时的。换句话说,“Wolfram/alpha承诺将每个人变成即时的万事通”。

 

那么Wolfram/Alpha是怎么做到的呢?

对于数据,它依赖因特网上现存的巨量信息和数据库。对于计算和可视化,它利用MATHEMATICA的能力。然而,说着容易做起来难。在今天(09428日)对外宣布之前,Wolfram Research公司雇用了100多个人,用了10年的时间才完成这一项目,两周之内(20095月初)将会公开上线。

Wolfram/Alpha有四个主要构成部分:

i)数据收藏——虽然大量的数据库存在于WWW网上,其中的大多数使用的都是不兼容的格式、不同的语言,有时则带有前后不一致和错误的数据。Wolfram/Alpha首先必须清理、关联、检查及核实这些数据库,并在应用它们前将其转变成一致而持续的格式。这是一项极其耗时的繁杂工作。

 

ii)计算机演算——因为MATHEMATICA开发得相当成功,这一部分相对简单。

 

iii)语言能力——虽然人工智能关于理解普通语言的问题仍然没有得到解决,但理解一个查询的问题,即使是不规范的,也能够被分解成有限的子问题组,并得到处理。我们在“自动电话应答”软件中已经见过,它出现在客服中,在许多厂商和设备中很流行。换句话说,我们知道当键入一个查询时,我们并不是要与电脑闲谈或询问它今天的心情如何。我们脑海里有着特定的目的。这使得理解自由形式语言相当简单。

 

 

iv)自动呈现——这与图形用户界面(GUI)和用户友好设计有关。同样这一方面也被很好地理解。

将这四种任务集合起来,这就是Wolfram/Alpha。创建者声称,到今天为止,它覆盖了一座一般参考书图书馆里95%的知识。当然,项目正在不断发展,越来越多的数据库和性能会被整合进系统中(类似GOOGLE90年代时它与现金与很大不同,能力也小得多)

 

听众的问题和回答

Wolfram/Alpha如何处理不一致、不完整和不确定的数据?回答:只要可能,W/A会提供原始资料、警告、脚注以及可应用于结果的不确定性范围。

 

Wolfram/Alpha是否有API文件?回答:会有的。当然,因为Wolfram Research是个商业公司,Wolfram/Alpha的一部分将会是私人所有。

你们是否能够个人化Wolfram/Alpha?回答:是的,只要你拥有了API入口。

Wolfram/Alpha的答复页面还有其它什么信息?回答:用于获取答案的推测和进一步查询的选择。

此次宣讲会放到网上吗?是的,会及时放上去。(请关注这一博文,一旦出现了我马上放上来。它很值得你花费一个半小时。)

我问自己的两个问题:

是什么让哈佛成为哈佛? 因为许多重要的发现和宣告在世界知道之前,其创建者本人就在这里进行了现场宣布。

我为什么写这篇博文?因为这样你们读者就可以说,我是首先在科学网上看到的。(梅进 译)



https://wap.sciencenet.cn/blog-1565-228880.html

上一篇:An American Success Story – living the American Dream
下一篇:Mathematical Game theory (#6)
收藏 IP: .*| 热度|

18 唐凌峰 张志东 王晓峰 刘全慧 阎建民 梁进 刘进平 曹广福 曹聪 杨秀海 周春雷 陈苏华 刘秀群 孙健钧 陈学锋 孙静宇 hypersurface wangyi6

发表评论 评论 (16 个评论)

数据加载中...
扫一扫,分享此博文

Archiver|手机版|科学网 ( 京ICP备07017567号-12 )

GMT+8, 2024-4-29 09:31

Powered by ScienceNet.cn

Copyright © 2007- 中国科学报社

返回顶部