The Relativity of GenAI Productivity Gains and the Jevons Paradox
In this article, I will explore the concept of the relativity of GenAI productivity gains and its connection to the Jevons Paradox.
GenAI is a rapidly developing field with the potential to revolutionize many aspects of our lives. One of the most exciting aspects of GenAI is its ability to significantly enhance our productivity. However, opinions differ on the extent of these gains. Some believe that GenAI will dramatically increase the work we can accomplish, while others believe that the gains will be more modest. I’m convinced GenAI increases productivity, but I also think those productivity gains will be absorbed by increased expectations. Let me explain.
I’ll use my background in IT as a starting point to explain my point of view.
It’s not the first time a new tool has been introduced to improve productivity. Remember the time when Integrated Development Environments (IDEs), application frameworks (like Spring), or code completion were introduced? Back then, these tools boosted our productivity, and we saw it as a natural progression, an evolution in our craft. What I don’t recall are questions about how fast we would go or any return on investment (ROI) analyses that needed to be made with the introduction of these tools.
My feelings about the introduction of Generative AI are different. It starts with the sheer amount of buzz that comes with its introduction. Moreover, a big part of the conversation around GenAI is unprecedented and focuses on productivity gain expectations and faster delivery. Expectations are often sky-high, and GenAI tools and usage are usually expensive. This frequently seems to result in investment calculations based on the tooling cost being countered with improved developer productivity. Wut?
The Relativity of Velocity and Speed
When people introduce terms like faster, velocity, and speed into these conversations, my first question is, “faster compared to what?” Speed is relative; to let it make sense in an expression, we must clarify its context and what we compare to. GenAI will undoubtedly allow us to accomplish more at the same time, but does this mean we will deliver things faster?
Could all GenAI improvements be consumed through improved quality, better security, cleaner code, or enhanced maintainability in the long term? Moreover, could our increased productivity be consumed by extra features, where we would deliver the same product in the same timeframe but richer in content? This last assumption is especially tricky as it yields value that is hard to measure, mainly because extra features in the same timespan could become the new “Standard”, a baseline shift.
Shifting baselines
Compare it to Formula One: every season, new cars are released that are faster in absolute numbers, but compared to each other, the changes are minimal. Comparing the relative speed between cars, there’s no noticeable change. The same could likely be true for the results of introducing GenAi in the day-to-day work of many crafts; productivity will rise, but that increased productivity will be equal to that of competitors.
The Jevons Paradox
This brings me to the Jevons Paradox. Here’s a snippet from the Wikipedia page describing this paradox:
“In economics, the Jevons paradox (sometimes Jevons effect) occurs when technological progress increases the efficiency with which a resource is used (reducing the amount necessary for any one use), but the falling cost of use induces increases in demand enough that resource use is increased, rather than reduced. Governments typically assume that efficiency gains will lower resource consumption, ignoring the possibility of the paradox arising.”
The Jevons Paradox tells us that when technology becomes more efficient, we tend to use it more, which can offset the gains in productivity. The way lighting became more efficient is an excellent example of this paradox, especially with the introduction of LED lighting, which is more efficient. However, this introduction also led people to install lighting everywhere and have it constantly turned on. Often this resulted in raised energy bills instead of lower ones. In the case of GenAI, the outcome of the paradox could be that we may be able to write more code and deliver more features, but we may also spend more time refining and perfecting them. Additionally, a new baseline for features that define a project or product a success could inject more extra work than GenAi productivity gains could yield, hence the paradox.
Conclusion
While GenAI offers promising enhancements to our productivity, it is essential to approach these advancements with a nuanced perspective. The initial excitement and significant productivity gains may be tempered by the increased expectations and the potential of the Jevons Paradox, where greater efficiency leads to increased usage and demand. By understanding the relative nature of velocity and productivity, we should be realistic about our expectations. Ultimately, the true measures or benefits of GenAI’s impact could be not in how much faster we can work but in the quality of the improvements it brings to a craft.