This weekend’s InsideAI newsletter included writer/venture capitalist/angel investor Rob May’s commentary on how AI could change the laws of economics. Economic laws change overtime, May writes, and AI could drive changes into economic laws by augmenting the importance of knowledge curve, changing the concept of demand curve, and introducing data-driven discontinuities. It was interesting to see how a seasoned practitioner of the supply and demand in the knowledge marketplace sees the laws of economics.
Sometimes, I think, the power of economic law is that it is so all-encompassing (so vague) that it can accommodate most changes that time brought. Really, Adam Smith cannot be vague-er that what he said about the forces of the market still largely hold true after a few industrial revolutions, a few world wars and some communist revolutions. In microeconomics, the same form that production function take can express utility; the national GDP also symbolises the aggregate demand in the macroeconomy… They all seem to “commute”. In addition, there are often schools of thought that argue with each other, the most famous being the Keynesian school and the Friedman school in matters of monetary economics, and of course, numerous others. Don’t like this expression of the economy? Don’t worry, you can always find another economist saying somewhat opposite things. Economics is so young, so frivolous, so shady, and so ambitious.
In seeing this post though, the foresight of economics rises to the thought.
Existing Knowledge and Discontinuity in Economic Models
The knowledge component needs not be introduced to supply functions, because it has been built into the most popular production function almost a century ago. The Cobb Douglas production function, developed in the 1920’s, incorporates the Total Factor Productivity (TFP), which models the knowledge, technology, or efficiency of a production unit in utilizing its capital and workers. The TFP could be dynamic and dependent on the levels of the means of production, the knowledge of the production unit itself, or some other firms or the economic environment. The efficiency of production of a unit can then be analysed using notions such as returns to scale.
This extend nicely to the third point mentioned by the commentary author on data-driven discontinuities on the mathematics
of economics. Mr. May raised a capturing point about the sudden elevation in performance given availability of data.
It may not look like it, but there are existing numerous discontinuities in individual- or firm-level supply and demand
functions, only smoothed over by aggregation or by assumption of continuity. If additional discontinuities on the
individual or aggregate level do occur as driven by data, smooth step or other sigmoidal functions could come to aid in
smoothing over them. The tipping point feature mentioned by the author sounded similar to phenomena such as the
network effect, in which a tipping point marks a jump in efficiency. This can be captured by notion of returns to scale
called by Hal Varian as the demand-side returns to scale in this collection of thoughts,
alongside other interesting phenomena.
Data As a Means of Production
Mr. May’s commentary has contributed to the vast scape of discussions on Artificial Intelligence. As for me, my personal interest is most stirred by data ownership in the age of AI: the fundamental means of production in capitalistic models are currently two: capital, and human capital. Which one do data belong to?
Do we own our data? We certainly do not collect or enable generation of many forms of the data without many of the technologies. For example, without maps and GPS services, we cannot keep track of where we go down to the degrees. However, we do move around generating data, and the the data describe us. Data, if a form of capital, can be owned by firms. This is currently not allowed by the legal framework in many countries. It cannot be a form of capital until a clear definition of ownership is defined.
Is data a form of human capital, then? Do workers own them, and how do workers offer them for rent/wages? Again, this is loosely defined legally, and in many cases not enabled by technology firms. For example, Facebook does not enable a reward platform for users to get monetary reward in return for privacy, but offers an exchange of service, which falls into the definition of a barter system. If workers do not own or be able to monetise data, it cannot be human capital.
Capital and human, the fundamental means of production in a capitalistic world. This definition has thrived for 200+ years. As the data economy develops, would we see the classification of data as new form of the means of production, introducing a new twist in economics? We just might. It is an exciting prospect.
Written by Natasha. Last edited:2020-07-27 02:36:14