A.I in Finance, in General.

As I finish up my official post for this week, I thought I’d share a really interesting and informative video on A.I in finance:

The video is done by Siraj Raval, a Youtuber with a whole pool of great videos on different implementations of A.I. I like this video because Siraj really clearly lays out the many shortcomings of trading companies which could be improved with A.I-based algorithms. For instance, I didn’t know that investment banks are trying to use technology which can recognize fraudulent stock orders, to improve security and avoid losses. Also, did you know that banks are testing A.I which can predict whether or not a customer will pay back their loans? And, did you know that 90% of the world’s data (not just financial data, but ALL data) was created in the past two years? I can hardly wrap my mind around the last one, but if it is really true, then this is probably the best time for testing A.I on financial data.
Up until now, my view of A.I in finance was pretty restricted to trading algorithms, but I know see that this is just one of the many (he brings up 5 categories in the video, with several subcategories to each) areas of interest.
What I found even more intriguing was that Siraj actually brings up the idea of using vector support machines to learn financial data. This is nothing groundbreaking, but it is interesting to consider how I might use VSMs in my work. Vector support machines are algorithms which can classify a data set of vectors into different categories, and then use this knowledge to classify future, unknown vector entries into these defined categories. Essentially, VSMs can predict where a vector ‘belongs’. This could be useful, as both Word2Vec and BERT generate vector representations of information. Whichever one I use, I might want a system which can classify my vectors in order to improve learning or efficiency.
Finally, one quote that resonated most with me was this:
“If you can tackle a single, niche problem very well, then you are golden. That’s when they [hedge funds] will come to you, and when you will build a brand around this.”
This really sums up [my understanding of] the current state of affairs in quantitative finance. Right now, there is so much data and so much technology available to investors and hopefuls alike, that there is no longer a point to trying to ‘predict the stock market’ as a whole. Rather, all this data means there’s an abundance of small, specific opportunities which can be capitalized upon, and everyone is competing to see whose specific opportunity yields the most return, most efficiently.

Leave a Reply

Your email address will not be published. Required fields are marked *