STAC's working definition of artificial intelligence (AI) encompasses both neural approaches like Deep Learning as well as non-neural machine learning approaches such as gradient boosted trees and genetic algorithms.
The financial markets were early adopters of AI, and its use is growing: growing in the number of organizations using AI techniques, the range of techniques being deployed, and the purposes they are putting them to.
STAC focuses on two aspects of AI:
- What is possible. There is ongoing--sometimes conflicting--dialog within the STAC Benchmark Council about use cases, including applying Deep Learning to trading, and whether AI is better for developing strategies directly or providing clues to human quants, and why some of the peculiarities of finance problems makes machine learning challenging. There have also been overviews of artificial intelligence as a whole and Deep Learning in particular, as well as specific cases, such as using Deep Learning to predict market movements or using FPGA to accelerate strategy development through genetic algorithms.
- How to make what is possible doable. Data scientists and data engineers face many decisions, from choosing techniques to choosing technologies to tying it all together in data science pipelines that best serve the business. STAC is now engaged in research in these areas, one example of which is a study on scaling properties of a natural-language processing (NLP) use case common in finance. A STAC working group is now forming around AI benchmarking. Please see the working group site if you're interested.
The growth of AI in financial markets is a significant driver for other hot topics within the STAC community, including quant finance as a whole, the role of cloud computing, and the storage and memory revolution.