The Past Speaks

Niall G. MacKenzie, Stephanie Decker, Christina Lubinski, three business historians, have published a short but excelent piece in the British Journal of Management that puts Generative AI into some historical context “Contested Imaginaries Through Time: Putting Generative Artificial Intelligence in Context“. Their piece, which uses on the theory of the German sociologist Jens Beckert (see his book Imagined Futures: Fictional Expectations and Capitalist Dynamics) cites a number of historical studies, including works on the Industrial Revolution era by Joel Mokyr and Mary O’Sullivan.

I don’t disagree with anything MacKenzie et al. say in their paper. However, I have parallel set of observations that are about the use of historical analogy in the marketing of OpenAI’s flagship product, ChatGPT. (When I say marketing, I’m talking more about the attempts by the company to secure the support of investors and policymakers than the end consumers). The very term ChatGPT, is an allusion to a term that emerged from the field of economic history in, if memory serves me correctly, the early 1990s. At that time, Joel Mokyr, Nathan Rosenberg, and other who study industrial revolutions past and present, popularized the term General Purpose Technologies (GPTs) to describe any technology that have the potential to significantly impact multiple sectors of the economy, leading to widespread and transformative changes in various industries. [It may be that Robert Solow first introduced the term but in any event it was in common currency in econ history circles by the team I reached university in 1995]. Steam engines and electricity are cannonical examples of GPT. Since the 1990s, generations of students in intro economic history have been taught that the key characteristics of General Purpose Technologies are:

Breadth of Application: GPTs have the capacity to be applied across a wide range of industries and sectors.

Economic Impact: They have the potential to bring about substantial economic growth and TFP improvements. They allow lots of people to produce more GPT per hour worked.

GPTs tend to facilitate and support the development of complementary technologies, further amplifying their impact on the economy.

Long Gestation Periods. The development and diffusion of GPTs often take a considerable amount of time due to their complexity and the need for widespread adoption. It is took a long time for the steam engines invented down in Cornwall to really change the world, but once they did, the social returns on the initial investments on steam engines were just massive.

The long gestation period teaching from economic history stresses the need for patient capital. Telling your investors that your venture is working on a GPT is basically a way of explaining in advance why it is going to take a long time for venture to generate real profits for the investors or society as a whole. I would be fascinated to hear from readers what actual Venture Capitalists think when this historical analogy argument is used on them. (Information in the comments section welcome!).

Now since the 1990s, millions of people throughout the English-speaking world, and probably beyond. have probably been exposed to the concept of the GPT through introductory economic history class or maybe a geography class. (I first encountered the term in a geography course). For years, I delivered a few lectures each semester to business undergraduates that were designed to give them the overview of economic history they needed. So they got some Great Divergence, some new institutionalism, and, yes, some exposure to the concept of the GPT. So have millions of learners around the world– at least judging from the textbooks that are widely used. So by 2022, lots of people were vaguely familiar with the concept of the GPT. They were a small proportion of the human race, sure, but they were disproportionately in positions of power and authority. I bet everyone who did a PPE degree at Oxford or spent a couple of years at Stanford, which is what Sam Altman did, hear about General Purpose Technologies. That gives elites a common vocabulary for thinking about the world.

AI Enterpreneurs Meeting With a Key Stakeholder. Historical analogies can give people a common vocabularly for thinking about the world.

24/05/2023. London, United Kingdom. The Prime Minister Rishi Sunak meets with Demis Hassabis, CEO DeepMind, Dario Amodei, CEO Anthropic, and Sam Altman, CEO OpenAI, in 10 Downing Street. Picture by Simon Walker / No 10 Downing Street

What all of that means is that it was a brilliant use of rhetorical history by Sam Altman’s team at OpenAI to label their product ChatGPT. Rhetorical history is when a company, manager, or some other actor in the academy uses a historical argument or allusion to persuade others. As someone who studies how companies use historical ideas to appeal to consumers, investors, and other groups, I am impressed by what a clever move it was to dub OpenAI’s product ChatGPT. I say that as someone who is agnostic about whether generative AI, as opposed to say AGI, truly is a GPT. (The late Nick Crafts had some interesting observations about that).