Fed Predictions: AI or Markets?

In a recent blog post, Vasant Dhar had some interesting comments about the potential role of AI in Fed policy:

Could AI help us better understand the data and design better economic policy interventions?

Last week I had lunch with some central bankers who wanted to talk to me about AI. They were keen to discuss how to think about the risks and opportunities associated with the increasing presence of AI in financial markets. I asked them the same question I asked Paul: could AI do the job of the central bank? Over the course of the meal, they realized that an AI would indeed have a lot to offer in terms of connecting the dots better than humans can. The dots include historical data on all past central bank monetary policy decisions, past communications and forecasts, all historical and contemporary data available at the time, and all literature on the subject over time published. In principle, we believed that AI could better model the economy. At the very least, the models should be compared with those currently in use.

But we still need people to blame, they concluded. After all, you can’t get mad at an AI when it’s wrong.

I certainly see the appeal of using AI in formulating monetary policy. But on balance I would prefer to rely on a market approach to policy.

Could an AI predict better than an NGDP futures market? That’s hard to say. An AI forecast can include market predictions, as well as other factors missed by market participants. In that case, you could think of AI predictions as human/machine hybrids. In the early days of computer chess, games that combined the insights of both grandmasters and computer programs were better than humans or machines working separately.

But if AI is truly superior to the market in some respects, we can expect market participants to use AI in trading financial assets. If so, insights from various AIs will be incorporated into market prices.

From this perspective, the addition of AI is not a qualitative change in markets or forecasts. Instead, the AI ​​revolution will add many really smart ‘entities’ to the market, making it even more efficient than before. This may be important, but it is still more of a quantitative than a qualitative change. Even if some AIs are smarter than any individual, including Jay Powell, no individual AI will be smarter than a market with many smart people and many more even smarter AIs. The wisdom of the crowd still holds up.

A potentially more promising use of AI would be to come up with the right Fed policy target (e.g. inflation versus NGDP growth, or levels versus growth rates). These questions cannot be answered by market forecasts because there is no point in what future data clearly resolves the question of who was right or wrong. Naturally, this applies to a wide range of public policy issues. So we might imagine asking AIs whether the death penalty increases or decreases the overall utility of society. If this thought makes you feel sick, you can take some comfort in it AIs are currently far from being able to answer those types of questionsand it is not clear that they will ever be able to do so.

P.S. David Beckworth asked an AI what he thought about the use of AI in monetary policy.

(1 COMMENTS)