Earlier this year, FS Insider discussed a new machine-learning “forecasting engine” developed by San-Diego-based Intensity Corporation used for economic and revenue forecasting, large-scale investing, supply chain optimization, and a wide range of other areas. The current forecast their platform is giving for a US recession is June 2019, which is updated daily on their website here along with a one-year forecast history (shown below).

Get The Timeless Reading eBook in PDF

Get the entire 10-part series on Timeless Reading in PDF. Save it to your desktop, read it on your tablet, or email to your colleagues.

AI Recession Forecast forecast performance

Source: Intensity Corporation

As originally mentioned, Intensity boasts a number of very large tech firms as clients—Apple, IBM, Microsoft, and others—and is itself comprised by a team of data scientists, statisticians, and econometrically-minded PhDs.

Though progress in AI continues to remain underappreciated by the broader public, almost every day a new discovery or breakthrough is announced, including the ability of AI, more recently, to figure out a Nobel Prize winning experiment all on its own in one hour along with endless defeats of humanity's best and brightest at our most challenging games.

With such monumental progress being made in the area of machine learning—a subfield of AI that gives "computers the ability to learn without being explicitly programmed," as Arthur Samuel defined it—the big question is, for those interested in tracking the US business cycle, can AI also do a better job than humans in predicting recessions?

As Intensity's researchers explained on our podcast in May, their platform has already shown this to be true. Here's how they put it:

Intensity’s state-of-the-art scientific testing and validation methodologies ensure superior short and long-range forecast performance. We carefully correct for biases resulting from spurious correlation and data snooping. We measure the accuracy of our forecasts, quantify uncertainty, and ensure optimal forecast performance at each point in time. In contrast, alternative approaches typically impose unreliable relationships among economic factors or rely upon the opinions of professional forecasters, who are susceptible to cognitive bias and unable to discern high-dimensional, time-varying relationships.

When we consider the traits identified on how to become a superforecaster, and the dreadful forecasting record of the Federal Reserve and other institutions relying on human-derived macroeconomic modeling techniques, it is interesting to think that AI may eventually run monetary policy at the Fed and have better judgment than most politicians.

Article by Financial Sense

Save