“Seldom do more than a few of nature’s secrets give way at one time. It will be all too easy for our somewhat artificial prosperity to collapse overnight when it is realized that the use of a few exciting words like information, entropy, redundancy, do not solve all our problems.”
Claude Shannon, known as the “father of information theory” and the “father of the Information Age,” was an American mathematician, electrical engineer, computer scientist, and cryptographer. In 1948, he laid the mathematical foundations of “information theory.” Eight years later, he cautioned against the overuse of his theory in an article titled “The Bandwagon.” (1) fermatslibrary.com/s/the-bandwagon
Shannon’s information theory, as popularized by Warren Weaver and John Robinson Pierce, fundamentally redefined “information” as a measure of potential rather than content. It quantifies the freedom of choice in selecting a message, focusing on what could be communicated rather than what is actually said.
In deriving the functional form of entropy, Shannon made several key assumptions: continuity (2) $H$ should be continuous in the probabilities $p_i$. , monotonicity (3) If all the $p_i$ are equal, $p_i = \frac{1}{n}$, then $H$ should increase with the number of possible events $n$. With equally likely events, there is more choice or uncertainty when there are more possible events. , and the composition law (4) If a choice can be broken down into two successive choices, the original $H$ should be the weighted sum of the individual entropies. . These are often taught in introductory courses on information theory. However, a more fundamental assumption that is frequently overlooked is whether we can define a probability space at all.
Defining a probability space requires establishing a well-defined sample space that encompasses all conceivable outcomes. Given the open-ended nature of human creativity and the boundless possibilities of our world, this is a formidable task—if not impossible. While getting into the set-theoretic and measure-theoretic details is beyond this post’s scope, the debate about the existence of a universal probability space touches on philosophical questions about humanity’s ability to comprehend the unknown unknowns. I believe in humanity’s infinite potential to create new possibilities that we cannot foresee today—a belief confirmed by history.
Even with this critique of applying entropy to measure AI’s potential to automate tasks and ultimately displace jobs, I cannot dismiss the relevance of entropy in understanding the new horizon before us. In the intelligence age, machines may become more intelligent than us, at least by our current definitions of “intelligence.” Entropy offers insights into where this undeniable machine intelligence could render human work unnecessary. However, to get a complete picture, we should also consider the notion of fault tolerance, as these intelligent machines are not verifiable systems and may never become so, given the nature of intelligence.
With this in mind, we can envision a future where intelligent machines undertake low-entropy tasks—tasks with less freedom of choice—that are also fault-tolerant. This shift may lead to the loss of many jobs, but it doesn’t necessarily result in mass unemployment. Economic mechanisms don’t operate in such a straightforward manner.
According to the “Tasks Model” framework (5) nber.org/papers/w31910 , automation technologies enable firms to substitute capital for labor across an expanding array of tasks, profoundly impacting employment and the economy. Automation leads to:
- Displacement Effect: Workers are pushed out of roles as machines and software systems take over routine and even complex tasks, reducing the labor share in adopting industries.
- Productivity Effect: Automation increases productivity and lowers production costs.
- Shift in Occupational Structure: Demand decreases for workers in automated tasks, potentially reducing real wages and employment opportunities for displaced workers if they cannot transition to new roles.
This dynamic underscores a critical intersection of automation, employment, and capital, highlighting the need for a workforce adept at adapting to roles that leverage uniquely human creativity and problem-solving skills.
So, how can we leverage this knowledge to our advantage?
The first step is self-awareness. Reflect on your own profession and examine the tasks you perform daily. Identify those that are repetitive, rule-based, or easily codified—these are the ones most susceptible to automation in the near future. Recognizing this reality is akin to analyzing your skill set’s investment portfolio.
Once you’ve assessed this, focus on developing skills that enhance your adaptability and long-term value. It’s not about avoiding automation but learning to work alongside technology and cultivating uniquely human abilities that are harder to replicate. Skills like creativity, complex problem-solving, critical thinking, effective communication, and collaboration become indispensable. Embracing a mindset of lifelong learning and adaptability is crucial in this ever-changing landscape.
By proactively shaping our own futures, we can navigate the era of automation and AI with confidence and optimism, turning potential challenges into opportunities for growth. It’s about becoming more entrepreneurial in how we think about our careers—always looking ahead, scanning the horizon for new opportunities, and constantly learning new skills.
Moreover, as a society, we must engage in open-minded discussions and experiment with solutions to ensure the benefits of automation are shared more equitably. Ideas like universal basic income, better worker retraining programs, and new models of ownership where workers have a stake in the companies benefiting from automation are worth exploring.
This isn’t just an economic issue; it’s a social and political one. It’s about how we choose to navigate this new era of automation and AI together. The future of work is not predetermined. By understanding these trends and actively participating in shaping the future we want, we can turn the tide in our favor.
In a world where order is increasingly handled by intelligent machines, embracing the disorder—our creativity, adaptability, and humanity—is the way forward.