We are cautious optimists on AI’s eventual productivity
impact. The lack of a measurable impact thus far is not
a good reason to downplay its potentially
transformative effect. Previous general-purpose
technologies have taken time to raise productivity.
Economic history suggests that job creation and
productivity enhancement from technological change
more than outweigh job destruction over the long run.
That said, one risk is that the scope of job types under
threat from AI means this time could be different.
Certain sectors will be outsized beneficiaries from AI.
In the near term, these are ‘enablers’ like chip
manufacturers, ‘scalers’ such as platforms, and ‘early
adopters’. In the long term, those with large numbers
of knowledge workers and lots of administration, such
as finance, education and the law, are the biggest
beneficiaries (from the perspective of capital).
Governments and regulators face a pacing problem
whereby the rate of innovation is so rapid that policy
struggles to keep up. We anticipate a wave of AI
regulation, focused on human oversight, accountability
of decision making, privacy, and bias.
Finally, AI hardware and software will become a new
locus of geopolitical competition. Export bans of
leading-edge graphics processing units are already
part of the arsenal of US-China rivalry. The values
embedded in AI decision making, and its dual-use
military and commercial applications, raise the
prospect of a cyber arms race.