I will return with my usual note on Monday. In the meantime, given the volume of client enquiries prompted by Citrini’s paper on the consequences of an AI-driven “abundance of intelligence” - and the market volatility that has accompanied it - a brief response seems warranted.
To begin with, any research that sparks this level of debate is valuable and should be welcomed. It is also the case that there is an enormous amount of uncertainty around the impact of AI - in such circumstances thought experiments, such as the one set out in the note, are useful.
Nonetheless, our reading is that the paper overstates the likely macroeconomic costs of AI. Its central weakness is that it approaches the issue primarily from a microeconomic perspective. The argument hinges to a large extent on the notion that AI will disintermediate processes to such an extent that whole firms - even sectors - could be swept away. The paper lays out examples ranging from insurance to travel agencies.
Yet what is disruptive for a particular industry is not necessarily detrimental for the economy as a whole. Removing layers of intermediation will reduce the revenues of firms providing that intermediation (i.e. the insurance companies or travel agencies). But it will also lower costs for the firms and households that were previously consuming these goods and services. The former group experiences a (potentially substantial) reduction in spending power as a result of technological innovation, but the latter experiences an increase.
At a macro level, the crucial issue is the balance between aggregate supply and aggregate demand. A plausible risk is that the benefits from the rapid deployment of AI accrue to a small number of tech firms in the form of excess profits and, if they seek to save these profits, this could cause demand and thus income to fall across the economy. The result would be a rise in unemployment and increase in deflationary pressure.
But such outcomes do not happen in a vacuum. Governments retain powerful tools. Taxing excess profits and redistributing them via public spending would be the most direct response. The pandemic demonstrated clearly that when confronted with severe economic shocks, policymakers are willing and able to act at scale.
AI will undoubtedly be highly disruptive. Some firms and sectors will contract, perhaps sharply, while others expand. This process will create losses for some and gains for others. It may also cause significant financial dislocation as those losses wash through the system. This is the historical pattern with general-purpose technologies. On balance, however, the aggregate economic effects are likely to be positive. For more on our AI research - including our updated AI Economic Impact Index - see here.
