The 2023 edition of the Journées de Recherche en Apprentissage Frugal (Frugal Learning Research Days) gave us an opportunity to gain some perspective on the reasoned uses of Artificial Intelligence (AI).
In recent years, computing power and Big Data have greatly developed AI programs. In the field of vision or decision support systems, AI contributions are revolutionizing our daily lives and bringing us a higher quality of life. However, Machine Learning (ML) or Deep Learning (DL) consumes enormous amounts of data to understand and learn to decide, not to mention the laborious work of the little hands who annotate this data. This “quality of life” comes at a price… Can we still afford it?
“A carbon footprint equivalent to that of 500 round trips by plane between Paris and New York. That’s the footprint associated with training AI models like ChatGPT for example.” Source TheConversation.com
Social pressure to excel and compete irresistibly pushes researchers, scientists (and everyone else) to ever greater levels of performance. Unbridled consumption, the race for gigantism, immediacy… Habits, incentives and society reinforce these injunctions. Yet the costs of producing and using these tools are rising exponentially. The impact on our environment is becoming unbearable in view of planetary limits. AI and technology can provide some solutions to postpone the deadline, but they are not the solution.
This time of abundance is over, given the finiteness of our planet’s resources, and everything calls for thrift and sobriety. Frugality is often defined as “doing better with less”. The “Keep It Simple, Stupid! or “Good enough” approach should also be considered, to optimize the use of our resources, but above all to reinforce sustainability through simplicity, adaptability and resilience.
What I took away from these days is that performance and robustness are not compatible, as @Olivier Hamant demonstrates. The quest for robustness facilitates homeostasis, frees our capacity to adapt and creates true frugality.
What can we learn from this for technology startups, their strategies and their management?
The race for performance drives startup growth strategy and team management styles. While high performance is essential to the creation and take-off of a technology start-up, this race is a marathon that is bound to last. In a rich environment and a growing market, the competition is open, and any optimized, highly specialized company stands a better chance. In a tighter environment and a market in recession, these qualities could stand in the way of agility and resilience.
Clarifying the vision, defining the missions and forging the raison d’être lay the foundations for sustainable growth. Faced with a world in perpetual evolution (VUCA) and accelerating, the startup that will last will be robust in technological, economic and ecological terms. And isn’t it said that creativity is born of constraint, and dies of freedom?
All this means that we need to rethink our value propositions, our business models, our market access strategies and the way we manage our teams. How can we do this? By changing our mindset and purpose. As a principle of simplicity, economy or parsimony, Ockham’s razor can frame this new approach. Let’s rethink performance in terms of robustness 😊 Designed to last!
#lowtech #AI4green #greenbydesign #startup4good