Generative Artificial Intelligence and the incommensurability of phase changes
There is some discomfort in the debate about Generative AI, with some wondering “where did all this come from” and others saying that there is nothing new.
But the fact is that since the launch of ChatGPT last November, use cases have materialized. As Paulo Silveira from Alura has said, the distinction from some other recent technological fads is that there is no longer a discussion about potential future repercussions, but about what is already being done in the present.
A useful analogy is the physical phenomenon of phase change in matter, where certain conditions are established and accumulate without a change in the state of the material — until it occurs suddenly.
The economy has been digitizing, with data being accumulated and made available, algorithms being developed, companies creating and testing business models, and underlying this, processing and communication capacity advancing logarithmically. Then, boom.
What is happening in Generative AI in recent weeks is the inflection point of a phase change that has been accumulating energy for at least 30 years — which, nevertheless, gives rise to a completely distinct and unprecedented reality.
This is a dynamic that has occurred before. Financial Times columnist Tim Harford analyzes the second industrial revolution and wonders why electric motors took 30 years to impact industrial productivity. The answer is that “the new electric motors only worked well when everything else also changed.”
But there is a characteristic of the digital technological regime that makes the current change more acute. The digitization of the economy has caused a sharp reduction in entry barriers, resulting in smaller companies having more relevant market positions. In the second industrial revolution, the transition point was using a huge boiler to power a factory, vis-à-vis, using electric motors and transformers. Now we have open-source projects, in public repositories, hosted in the cloud, carried out by enthusiasts.
All of this indicates that we will see a faster and deeper change than our intuition leads us to believe.
This perspective helps not only to bring some order to current events, but also to bring some level of humility to technology prophets.
There are several approaches to analyzing what is happening, but the common point of all is that, like in scientific paradigms, it is epistemically impossible to know what lies ahead. The different paradigms are incommensurable.
From Knightian uncertainty, to the dynamics of complex systems, with their emergent properties, and even the unintended consequences in attempts at regulation and control, it is simply not possible to understand a transformation that alters the foundational axioms of the system being analyzed.
We are entering uncharted territory.
