Accelerating the Quest for Alpha with AI

Extract from Cutter Business Technology Journal (Joseph Byrum)

Creating Market Insight with Textual Analytics

AI systems are dynamic, implementing a continuous process of observing data. For example, they can use natural language processing to absorb written reports and news items that often contain information that is essential to insight-driven businesses like the financial industry. Imagine how much more effective financial analysts and portfolio managers could be if they could enhance their understanding by surveying social media and other rich sources of business information in real time, drawing out the key developments and trends that would inform their choices. AI can also prioritize information flow so that it does not become overwhelming for the investment manager.

No human could possibly read everything, but a machine can. The best you can do with old-fashioned methods is to hire experts to sample a subset of the relevant data and produce written market insight reports. While these studies can be extremely useful, the resulting analysis is constrained by the amount of data sampled. Many companies already use forms of automation to sort through the data with machines, but, in the end, humans still have to read the data and decide what it means. Moreover, AI can help orient the analyst toward discovering what the information really means. Expert systems use cognitive engines that examine why something is happening, presenting possible scenarios to the analyst who will ultimately take the options presented and decide what to do.

The Defense Advanced Research Projects Agency (DARPA) Pilot’s Associate program10 from the 1990s was the first expert system designed to go beyond data collection to assist with decision making. This wasn’t an autopilot system. The idea behind the project was to take all the low-level decisions away from the pilot, so that the pilot could focus attention on the higher-level tasks that require human expertise. The system had four components that (1) reported on the status of the aircraft systems, (2) provided an assessment of the external environment, (3) helped plan the mission, and (4) assisted with short-term tactics to fulfill the mission.

The system could, for example, find the nearest airbase within range for landing after the aircraft spent longer than expected performing the mission. Rather than producing a list of every airbase, it would compute all the relevant factors and propose the “best” one — relieving the pilot of the burden of figuring all that out in the middle of a stressful situation. More importantly, it could determine the best path and maneuvers needed to avoid an enemy missile. It did so by making inferences based on knowledge baked into the programming by expert pilots.

Ground-based simulators proved that the technology worked, but DARPA realized at the time that the system envisioned was way ahead of what software and hardware could deliver in an airworthy package. That situation has finally changed, and AI will now be baked into the next-generation of fighter aircraft with the explicit goal of increasing the speed of closing the OODA loop.

In finance, hardware and software are finally powerful enough to fulfill the promise of augmented intelligence. AI offers an information-based, scientific approach that constantly reevaluates and assesses the situation so the investment manager is ready to achieve alpha by taking advantage of the opportunities that others miss. The system leverages the composite of all company, product, and customer experience data, including all unstructured and structured data. It then automates the process of delivering insights by identifying the
“what and why” that underlies behavior.