Algorithms in capital markets trading are changing shape. Watch out.

In this clear insight on trends in the capital markets, Yallop notes the recent switch to machine learning algorithms. It is not as if algorithmic trading hasn’t been around in the markets for a while now. But so far, it has been rules-based or, put another way, coded as an expert system. The disadvantage about such an architecture is that it does not learn. Rather, the expert, ie the programmer, has written down all his expertise and knowledge in the form of a code. This is then deployed in trading on the markets. The advantage is that, when mistakes happen, ie when the algorithm causes a particular market to malfunction, you can have a go at figuring out where the mistake lies. Machine learning algorithms are the reverse. They learn, which is great. But they are also a black box, because they have coded themselves. So, when something goes wrong, it is near impossible to figure out how it went wrong and how to correct for the future. That’s just one of the drawbacks. Yallop points out some more. It is not just social media that needs new forms of regulation.

Link to article:

You may also like to browse other AI articles: