Robots are on their way to cracking one of the world’s toughest codes: central banker speak.
In a matter of seconds, machines that mimic the human brain can read through dense and obscure policy statements and then offer a prediction. The humans who develop and use them say artificial intelligence gets it right, more often than not.
Robots that learn as they go “not only analyze communications faster than humans, but also mitigate several human shortcomings,” said Evan Schnidman, the founder of St. Louis-based Prattle Analytics LLC, which develops and sells computer-generated research on 15 central banks to hedge fund clients on Wall Street. “Human confirmation bias can lead to substantial analytical errors.”
For central bankers who have long thrived on wielding the power of words, growing use of the technology means they’ll need to pay even closer attention to their language choices, especially if computers seize on historical patterns humans tend to miss. Some central banks are already starting to vet communications through machines to gauge how they’ll be interpreted, although Prattle wouldn’t reveal which ones.
Schnidman, who did a PhD at Harvard University on how central bank communications impact financial markets before starting Prattle in 2014, charges $60,000 a year for three to five users to access analysis.
It takes about 45 seconds for the neural network of Prattle’s robots to read a 500-word statement and map the words over 80 billion connections to learn how the language is interconnected. It then draws on all prior language from that central bank to determine the likely market impact. For Federal Reserve minutes, it’s even faster: clients start receiving analysis in less than a millisecond.
This speed is one of the key reasons why advances in artificial intelligence are upending an area of research that, until a few years ago, might have seemed impossible to do without human common sense.
That’s because the task requires not only swiftness, but creativity. Policy makers have historically been hard to understand, partly because of the complexity of the topic and sometimes deliberately. Former Fed Chairman Alan Greenspan said in 1987 he’d “learned to mumble with great incoherence.”
These days central bankers are, generally speaking, trying to make their communications more transparent for a broad audience; on becoming Fed Chairman, Jerome Powell said he would attempt to communicate in “plain-English’’ and doubled the number of press conferences.
Policy makers have long been aware that their words can be as powerful as their actions for influencing markets. Bank of England Governor Mervyn King illustrated this in 2005 with his “Maradona theory of interest rates.” He was inspired by Argentinian soccer player Diego Maradona, who in the 1986 World Cup dumbfounded five English players to score despite running in a straight line.