Artificial Independence

Two things are hot topics in Washington D.C. these days: artificial intelligence (AI) and the Federal Reserve. Just Tuesday an announcement was made about the newest Anthropic model, Mythos, getting into the hands of nefarious actors. This is an AI model which has discovered many cybersecurity vulnerabilities across many platforms which were previously unknown. Also on Tuesday, Federal Reserve Chair nominee, Kevin Warsh, testified in front of the Senate confirmation hearing that he would not be a sock puppet if confirmed for the role but would instead maintain independence.

At its most fundamental level, artificial intelligence is really a sophisticated algorithm; a complicated set of rules that churn out answers to various queries. Couldn’t something like this be developed for interest rate policies? After all, we’re already hearing AI is taking over entry-level white-collar jobs. Why shouldn’t interest rate policy makers suffer a similar fate?

As it turns out, Artificial Independence has been proposed, studied, followed, and tossed into the circular file at the Eccles Building. John Taylor is an American economist and professor at Stanford University, known for his work on monetary policy. He introduced his set of rules known as the Taylor Rule in 1993 in a paper called “Discretion versus Policy Rules in Practice.” He wanted a simple rule that linked interest rates with two things: inflation and how well the economy is performing relative to its potential (i.e., the output gap).

Above is a chart from the Federal Reserve Bank of Atlanta (click here to see the article).  It shows various versions of the rule and where each version would place the overnight Fed Funds Rate as of February 2026. As you can see, all of them suggest monetary policy was too loose less than two months ago. Currently, the Fed Funds Rate target is between 3.5% and 3.75%, shy of the lowest mark of 4.28% from various calculations of the Taylor Rule.

To say following this rule would have produced better/worse results is counterfactual. We’ll never know. But one thing we can see is that February 2022 when inflation was the highest in a generation and the Fed was keeping its overnight lending rate nailed to the floor, these models called for the rate to be between 7.73% and 9.8%. Rules like the Taylor Rule do not eliminate uncertainty, but they do expose errors and impose discipline. If policy makers want to maintain independence, they may want to impose rules that support this notion.