# Campbell Harvey教授对于中国读者的答复

(darcylike) #1

1）大部分量化基金都是骗人的

2）基金经理将会被机器代替

【注：百度搜索“伍治坚+文章名字”可以找到这些文章。】

1）知乎用户：王勇

Harvey教授回复：我提倡3西格玛，而不是5西格玛。其原因正如您所说。

【原文回答：I do not advocate 5 sigma for exactly the reason you mention. In particle physics you have the luxury of a huge amount of data. We do not have that unless you are working at ultra high frequency level where you can credibly have billions of observations. The high frequency data, however, is very noisy.】

2）知乎用户：夏瑜

Harvey教授回复：我在2017年1月美国金融协会（AFA）主席演讲（《金融经济的科学未来》）中专门提到这个问题。

【原文回答：See my Presidential Address, The Scientific Outlook in Financial Economics and the discussion of August Compte. I agree with you that economics is, in many ways, more complicated because you are modeling interactions with humans. My comparison to physics was motivated by exactly your point – it is different.】

3）知乎用户：该隐

Harvey教授回复：我不同意。

【原文回答：I disagree. When I started my career, the fastest computer was the Cray-XP2. It was the fastest computer up to 1990 and did 1.9 GFLOPs (1.9 billion operations per second) and cost $25 million USD. It weighed over 5,000 pounds. The recently released iPhone 8 does 600 GLOPs and costs$800 and weighs 5oz. In contrast to most things in finance, forecasting the future of computer speed is very easy. We will soon reach the point where \$1000 buys you the computing power of a human brain. What is scary is that 5-10 years later, that same money can buy you the equivalent of a billion brains. This is exactly why Elon Musk is starting NeuralLink. Our brains, soon, will be greatly inferior to cheap machine.】

Harvey教授回复：上述观点在今天来说，基本正确。但是，在机器学习和深度学习领域，该观点不一定适用。

【原文回答：Yes and no. The statement is largely true today. However, there are important exceptions in the field of machine learning and deep learning. Here you simply state your objective and provide the data. The machine learns independently of the researcher. This is the whole idea of AI.】

5）雪球用户：不要数字只要胜利

Harvey教授回复：要想把人类的想法和行为归入模型中，是非常困难的。但是，基于一些典型的人类非理性行为，可以设计出一些最优秀的量化投资策略。

【原文回答：Agreed it is very difficult to model humans. That said, some of the irrationality leads to some of the best quantitative trading strategies. For example, we know humans tend to chase performance (buy when assets go up and sell after they go down). A simple algorithmic trend following algorithm can capture excess returns based on this.】

【注：伍治坚和Harvey教授的访谈录音（英语）在喜马拉雅FM/蜻蜓FM“伍治坚证据主义”播客栏目下。】

![](data:image/svg+xml;utf8,<svg%20xmlns=‘http://www.w3.org/2000/svg’%20width=‘750’%20height='217’>)