A rules-based central bank would be a major win
When I first heard about the idea of a rules-based Fed I scoffed, like most people. To my thinking, using a simple rule like the Taylor Rule would be stupid because it would miss all the nuances that go into analyzing the economy and setting rates.
I've changed my mind.
The thing is, it can't be a simple rule. It needs to take into account all the data available. It needs to be a Watson-level supercomputer that looks at thousands, if not hundreds of thousands of inputs in real-time. If the algorithm can handle news and politics, add that too.
What makes it such a compelling idea is that it could act quickly (why not set rates daily?) and diminish volatility. The value of reducing volatility and uncertainty is enormous. If you could give markets more certainty and confidence while lowering volatility, it would be a huge win.
There is no important central banking function a machine couldn't do. Aside from setting current rates, it could also send signals about future rates in much of the same way as the Fed. Along with daily rates, it could publish 1, 3, 6, 12-month forecasts and even further out.
Developing it and building enough trust could be a process over many years where it's rolled out as an input, slowly refined, turned into some kind of grey box and then used fully. If we're building computers that can diagnose diseases and drive cars, then it's time to get started on interest rates.
The fringe benefit would be that policymakers there would get a taste of the automation job uncertainty facing millions. That might even help them set policy in the meantime.