Posts from January, 2010
No comment yet
January 30th, 2010

Enlightened. Suddenly feel everything I’ve learned in the past becomes meaningful. On somthing BIG.

No comment yet
January 28th, 2010

没错,今天有两件大事:1. Apple发布了iPad;2. Armor Plus: Tekknoman Blade发布了。喜欢iPad的价格,大小,厚度。但是,直觉不喜欢一个全在苹果控制中的世界。每个公司都带有自己的个性的。这么多年了,Eric还是想他那个Workstation-Server的梦;Bill还在卖他的软件,因为“软件是需要钱来开发的”;而Steve还想他的控制整个生产链的世界。理性上知道通过控制终端,Apple的开发者可以赚到更多钱,但是作为一个封闭终端,是不是有点大?我不知道,只是反直觉。

No comment yet
January 23rd, 2010

用重典。钦此。

No comment yet
January 21st, 2010

Having been a month on the “new” point detector idea, I cannot get progress. Giving up and using FAST + Multiscale + Orientation Prediction + WLD.

No comment yet
January 18th, 2010

才能设计出这么繁复得清新脱俗,让人呕吐不止的API? http://code.google.com/p/facebook-actionscript-api/

No comment yet
January 14th, 2010

从昨天的地震之后,我觉得现在有必要根据手边的消息,用中文说说这事了。首先澄清一个信息,关于GFW试图通过MITM方式攻击SSL以获取Gmail内容的故事我是把通过一些渠道看到的消息整合起来作为背景信息提供给一个论坛,本意是想抛砖引玉,引出更具体的消息,因为这毕竟是件老消息了,没想到在Twitter上引起了二次传播,这并非本意。也就是委婉地说,这则消息是不可靠和不准确的。

首先,Google的声明占到了道义制高点,树立了自身保护用户隐私的公司形象,当然,这一形象对其后来发展会是至关重要的,作为一个广告公司,靠卖匿名化过后的用户隐私谋生,如果说用户隐私保护不当,那么就是一件太震撼的丑闻了。但是这不是故事的全部。

首先,一个商业公司,如果真的有上千亿的市场在前头,那么是断然不会抛弃利益而选择道义的。Google是一个商业公司,那么疑问就变成了,对于Google,中国是否一个具有巨大潜力的市场。Google已经做出了自己的判断。

许多人会用神奇的中国公式来论证Google是多么的荒谬,13亿的人口,乘以一个常数,就是滚滚的利润。但是Google不是卖糖水的企业。糖水可以卖给穷人,也可以卖给富人,但是知识产品没办法卖给什么都没有的人。印度有10多亿人口,但是你有牛肉市场么?多年的计划生育导致了一个畸形的老龄化社会,如果一个人均收入排名50名开外,却有一半的人口是老年人和少得可怜的年轻人,这样的社会,有什么潜力可言?

其次,Google的生意决定了有一些东西的破坏是会损害产品的核心价值,而这些核心价值的损害,可能会导致整个生意长期来看蒙受重大损失。而Google的排名算法就是他现在唯一赚钱业务的核心价值。删除一些能检索到的内容,没有问题,毕竟也屏蔽作弊的网站。但是要得寸进尺地要求能根据有关文件修改排名,将又红又专的网站提到前面来,那已经超过了这一生意的底线。一个靠排名生存的企业是断然不可能接受这样的条件。

而网上的舆论,要么是觉得国外企业竞争失败,找面子退出(特别点名批评Sarah Lacy同学:不识庐山真面目,只缘身在此山中。);要么是觉得Google和美国政府臭味相投,共同阴谋颠覆;或者说Google炒作,赚取眼球。而忽略了Google权衡的代价和得失。这是一个没有惊喜的理性决定,而这个决定却暗示了一个相对悲观的预期。

No comment yet
January 13rd, 2010

Someone has to stand up and say: enough is enough.

http://googleblog.blogspot.com/2010/01/new-approach-to-china.html

No comment yet
January 13rd, 2010

There are several practical ways to bypass GFW today. However, there are a good chance that all these services will be blocked in the future.

  1. HTTP/Socks Proxy: it is possible to block all proxies. First, have a black list of all proxies on the Internet. Second, block several default proxy port: 8080, 8000, 3124 etc.

  2. Tor: it is possible to block Tor directory sever. And the speed is too slow for practical usage anyway.

  3. VPN: it is possible to block most of them by blocking all IPs from Slicehost, Rackspace, Linode and Amazon EC2 (will cover large part of them). It is possible since China has blocked all Dreamhost IPs for sometime.

4.Proprietary Software: keep blocking their central severs.

Then, China can white-list all famous (censored) sites. So, yes, it is practical to censor the Internet and block every possible way to bypass GFW.

No comment yet
January 12nd, 2010

It has been a while that when people talking about general purpose image retrieval, it has to be bag of words model related. No doubt bag of words is a powerful model to solve image retrieval problem, however, is it the last model we have to solve the problem after all?

I don’t think so.

No comment yet
January 7th, 2010

[PDF]

Highlight: using reference of parts to carve the characteristic of a face.

No comment yet
January 1st, 2010

Fallacies: ad hominem, circular reasoning, composition, division, equivocation, appeal to popularity, appeal to force, accident, amphipoly….