Ke Jie was once the world's best player of the most complex game ever invented, an ancient Chinese board game called Go. But, in 2017, Ke was beaten by a computer program that had taught itself how to the play the game Ke had spent most of his life mastering.
The robotic victory marked a watershed moment for artificial intelligence (AI) and machine learning technology—a subset of AI whereby a computer learns to perform a task without being explicitly programmed to complete it, instead learning and improving from experience.
Now the technology is being applied in industries from transport, where algorithms are being used to teach self-driving cars how to navigate busy city streets, to health care, where robots are learning to diagnose and treat patients. And in finance, increasingly, these technologies are making decisions about what stocks to buy and sell.
Marcos López de Prado has been at the forefront of machine learning innovation in finance. The New-York based Spaniard was the first-ever head of machine learning at AQR, one of the world's largest hedge funds, before he left earlier this year to start his own firm, which sells machine learning expertise and algorithms to Wall Street.
The finance pioneer literally wrote the book on the use of machine learning in investing (his 400-page textbook Advances in Financial Machine Learning is included in the curriculum of several graduate school courses), and he was named the 2019 "Quant of the Year" by the Journal of Portfolio Management.
TIME sat down with López de Prado when he visited Hong Kong recently. Here's what one of the world's top quantitative analysts has to say about how robots are taking over global financial markets, and his great hopes for the technology.
Machine learning has become very successful at various industry applications beyond many people's expectations. A few years ago, we start seeing all these breakthroughs, like machines defeating the best players at [the board game] Go.
What has changed is that we have computers that are powerful enough to solve these problems. And we have a lot of data that before was not available and we know how to use this data in combination with algorithms.
So all of a sudden, machine learning algorithms have become experts at tasks where five years ago, only humans could be experts.
Finance is special because we do not come up with very large data sets compared to the data sets that are used to train these algorithms for things like recognizing faces or driving cars.
And markets evolve. You are an investor and you extract money from the market and the market learns to prevent you from extracting profits next year.
Read More: Artificial Intelligence Is Powerful—And Misunderstood. Here's How We Can Protect Workers
This oracle approach, this black box approach, is doomed to fail. What does not make sense at this point is to use these techniques to make black box predictions.
Unless you understand the reason why the model is working in the first place, you don't know when to switch it off.
Machine learning should be used as a research tool, not as a forecasting tool. It should be used to identify new theories, and once you identify a new theory, you throw the machine away, you don't want the machine.
In my view, this is the right approach in finance. Everybody should use machine learning to develop theories and test theories, but once the theory is uncovered, you should run the machine, the machine shouldn't run the theory.
Today most transactions are performed by algorithms. There are very, very few transactions that are executed by humans. Humans have been expelled out. Anyone who doubts the efficacy of this and machine learning in particular should look at the execution.
There is no reason to not expect that eventually these machines will be able to solve many tasks that we cannot solve, it's a matter of when, it's not a matter of if.
I think eventually, machines will take over. When is anyone's guess. Making decisions is about information and processing information. If you have a machine that can process information better and faster and more objectively, then they will succeed.
I think that eventually it will have a very positive effect in society. Assets are allocated more efficiently, the companies that are more likely to be successful receive the assets, the companies that are just not very likely to be successful, they will not receive assets.
My hope is that when these technologies are deployed correctly, they will make markets more efficient so when people invest, they will invest following scientific approaches rather than making wild speculation or wild guesses, and hopefully that helps society in the sense that, you want something as important as finance and investing grounded on scientific evidence not on sheer speculation.
AMY GUNIA / HONG KONG