Key Moments
A 100-Year-Old Solution to AI Work Slop
Key Moments
AI tools are rapidly advancing, but their practical workplace impact is uneven because we measure "efficiency" using arbitrary, unbounded metrics instead of true value creation.
Key Insights
The concept of "efficiency" evolved from a strictly defined, physics-based ratio (useful work divided by potential work) in the 18th and 19th centuries to the "arbitrary efficiency" used today, which is a ratio of loosely related metrics with no theoretical upper limit.
John Smeaton's 1750s experiments on water wheels established the initial technical definition of efficiency as a ratio of useful work to potential work, measured by weight lifted over a distance.
Thermodynamics in the 19th century allowed for the quantification of energy in heat and mechanical work, leading to the efficiency of steam engines being defined as output energy divided by input energy (heat from fuel).
Walter Polakov's 1921 book, "The Mastering of the Production of Energy," distinguished between "limited efficiency" (physics-based) and "arbitrary efficiency," which is based on chosen ratios and lacks an upper bound.
The "AI boomeranging" phenomenon, where companies rehire workers after AI implementations fail to meet expectations, highlights the disconnect between arbitrary efficiency metrics and actual value creation.
Shifting from "arbitrary efficiency" to "effectiveness"—measuring the quantity of valuable things produced—is proposed as a solution for knowledge work, focusing on outcomes rather than process-based metrics.
The puzzling gap between AI capabilities and workplace results
The current AI revolution presents a paradox: while AI tools exhibit rapidly expanding capabilities, their practical impact in the office remains uneven. Reports, such as one from Deloitte, indicate that most companies implementing AI haven't seen a significant acceleration in employee performance, leading to fluctuating ROI. Goldman Sachs, for example, recently reduced access to advanced AI coding tools, citing cost and productivity concerns. This discrepancy raises a crucial question: why aren't increasingly impressive AI tools translating into expected workplace improvements? The answer, according to Cal Newport, lies in understanding the historical evolution of a seemingly simple concept: efficiency.
The scientific origins of efficiency: John Smeaton and the water wheel
The modern concept of efficiency can be traced back to the 18th century with British engineer John Smeaton. His experiments on water wheels, documented in the Philosophical Transactions of 1759-1760, laid the groundwork for quantifying mechanical performance. Smeaton devised a method to measure "useful work" by calculating the weight lifted and the height it was raised. He then compared this to the "potential work" or "latent work" that could theoretically be done by the falling water, based on its weight and the distance it fell. This comparison yielded the first technical definition of efficiency: useful work divided by potential work, expressed as a ratio or percentage. The goal was to maximize this ratio, indicating how effectively the water wheel converted the water's potential energy into mechanical action.
Thermodynamics and the steam engine: quantifying energy conversion
The Industrial Revolution, particularly the advent of the steam engine, necessitated a more robust understanding of efficiency. Converting fuel into mechanical motion was complex, involving heat, steam, and intricate machinery. The development of thermodynamics in the 19th century provided the scientific framework to bridge the gap between heat and mechanical work. James Joule's experiments established the equivalence between mechanical energy and heat, leading to the unit of energy, the "joule." This allowed for the efficiency of steam engines to be defined as the mechanical energy output divided by the thermal energy input from the fuel. This precise, physics-based definition of efficiency, always capped at 100%, was crucial for technological advancement.
The shift to arbitrary efficiency in the early 20th century
As the 20th century dawned, thinkers from various fields—sociologists, progressive politicians, and business managers—sought to apply the rigor of scientific efficiency beyond physical machines. They wondered if human labor and societal programs could be measured with similar precision. This led to the emergence of "arbitrary efficiency." Walter Polakov, a Marxist engineer, critically examined this shift in his 1921 book, "The Mastering of the Production of Energy." He distinguished between the physics-based "limited efficiency" of machines and a new, more nebulous concept: "arbitrary efficiency." This new form of efficiency involves taking two loosely related metrics, creating a ratio, and calling it efficient without a strict theoretical limit or physical connection between the elements being measured.
The dangers of arbitrary efficiency in knowledge work
Arbitrary efficiency, while offering a way to quantify complex systems, carries significant risks. Firstly, because there's no upper limit, it can create unrealistic expectations and a perpetual drive for more, regardless of actual value. Secondly, the chosen metrics might not truly capture the system's dynamics or desired outcomes, leading to "spurious correlations" and potentially harmful decisions. For instance, a university might prioritize "graduates per faculty member," inadvertently encouraging the dismissal of teaching-focused faculty, thus degrading educational quality while artificially boosting the metric. In knowledge work, arbitrary efficiency often manifests as "tasks completed per employee." This metric is problematic because visible tasks (like replying to emails or attending meetings) are often only loosely correlated, or even negatively correlated, with actual value creation. The drive to increase this ratio, amplified by digital tools, leads to a "theater of knowledge work efficiency," where busywork is prioritized over meaningful output. AI, by automating visible tasks at an unprecedented scale, exacerbates this problem, leading to inflated metrics without corresponding increases in real value or profit.
Effectiveness: A 100-year-old alternative for knowledge work
The historical alternative to efficiency, which fell out of favor as machine-based work dominated, is "effectiveness." As defined by Jennifer Alexander, effectiveness is simply a measure of the quantity of valuable things produced. It doesn't concern itself with the *how*—the waste of resources or the specific process—but solely on the *what*: the valuable output. While efficiency is crucial for energy conversion in machines, effectiveness is more relevant for complex, messy endeavors like knowledge work, where human creativity and insight are paramount. The challenge is that measuring effectiveness is difficult, often requiring narrative and judgment rather than simple ratios. Cal Newport proposes a return to effectiveness by managers regularly asking knowledge workers to present their valuable contributions through storytelling and evidence. This approach shifts focus from metrics like "tasks completed per employee" to actual impact and value creation, providing a more authentic measure of performance.
AI's role: accelerating problems or enabling solutions?
The current trajectory of AI in the workplace, driven by a focus on arbitrary efficiency metrics, is problematic. AI tools are often used to amplify metrics like "tasks completed per employee" rather than to genuinely increase the production of valuable outcomes. This leads to situations where AI automates busywork, inflating statistics without improving the company's bottom line or product quality. The "AI boomeranging" phenomenon, where companies rehire staff after AI initiatives fail to deliver expected results, underscores this issue. However, AI could be a powerful tool for effectiveness if reframed. Instead of measuring code lines generated, AI could help individuals discover genuinely better ideas and solutions, as exemplified by Jaron Lanier's custom AI tool for materials science innovation. The key is to align AI development and deployment with the goal of producing meaningful value, rather than merely optimizing arbitrary, unbounded metrics.
Revisiting education and digital distraction
The discussion extends beyond the workplace to societal issues, including the impact of technology on education. A Stanford study on school phone bans revealed that while reducing phone usage improved student well-being, it didn't significantly impact test scores or attendance. Newport argues this is not a failure of the intervention, but rather an unrealistic expectation of a "simple solution." The positive impact on student well-being should be celebrated, as it addresses real issues like distraction and social pressures. Similarly, research on social media's effect on children suggests it's not mere "moral panic" but a genuine concern with observable harms, such as decreased attention spans and self-reported addiction. While scientific understanding of the neurological impacts is still developing, the social science evidence, coupled with adolescents' own admissions of unhappiness, indicates a real problem requiring serious consideration and intervention, such as school phone bans and potential social media restrictions.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Books
●Concepts
●People Referenced
Efficiency: Limited vs. Arbitrary
Data extracted from this episode
| Feature | Limited Efficiency | Arbitrary Efficiency |
|---|---|---|
| Definition | Ratio of useful output to maximum possible output (e.g., mechanical systems). | Ratio of two loosely related metrics, often without a physical basis. |
| Theoretical Limit | Has a maximum of 100% (or 1). | No theoretical upper limit; can potentially increase indefinitely. |
| Basis | Rooted in physical laws (e.g., thermodynamics). | Invented or chosen based on perceived relevance or ease of measurement. |
| Application | Mechanical engineering, energy conversion. | Knowledge work, management, social programs. |
| Potential Pitfalls | None inherent to the concept itself. | Measuring false correlations, creating unrealistic expectations, encouraging detrimental behaviors to boost metrics. |
Common Questions
The video argues that many AI tools are designed to optimize 'arbitrary efficiency' metrics (like number of tasks completed per employee) rather than true value creation. This focus on easily measurable, but potentially misleading, metrics can lead to busywork and doesn't necessarily translate to real-world productivity gains.
Topics
Mentioned in this video
A firm that published a report indicating that most companies implementing AI tools have not seen an acceleration in employee performance, with varied ROI.
An AI-powered CRM system designed for sales teams, featuring a 'pre-meeting summary' to streamline deal discussions.
Recently reduced the number of employees with access to advanced coding tools like Claude Code, citing cost and productivity impacts.
Mentioned as a company that has announced workforce reductions, citing AI.
Mentioned as a company that has announced workforce reductions, citing AI.
An online coaching program that pairs clients with a trainer for personalized nutrition and exercise plans, emphasizing daily accountability.
Mentioned as one of the companies citing AI when making job cuts, though the speaker suggests this is often to redirect funds to AI investment rather than actual replacement.
A local bookstore where signed copies of Cal Newport's new book can be purchased.
Mentioned as a social media app that teenagers use, and which parents may want to restrict access to.
Walter Polakov's 1921 book that introduced the concept of 'arbitrary efficiency' and distinguished it from the 'limited efficiency' of mechanical systems.
A scientific journal that published a review article discussing preliminary data on social media bans for children.
The publication where John Smeaton published his paper on water wheels in 1759-1760, detailing his experiments and the concept of useful work.
A 1972 book by Stanley Cohen that examines the concept of moral panic, often applied to new phenomena perceived as threats to society.
Publication that featured an article on new job numbers from the Bureau of Labor Statistics, discussing whether AI is taking office jobs.
The speaker's new book, set to be released on March 2nd, which focuses on systematically applying principles to create a meaningful life.
An advanced coding tool that Goldman Sachs recently reduced employee access to due to cost and productivity concerns.
A digital communication tool mentioned as an example of technology designed to increase arbitrary efficiency metrics rather than true value creation.
A video conferencing platform cited as a tool that contributes to fragmented schedules and increased arbitrary efficiency metrics, rather than genuine value creation.
An AI model mentioned in contrast to more effective AI tools, where the metric of usage (tokens, lines of code) does not necessarily equate to better solutions.
An 18th-century British engineer who conducted crucial experiments on water wheels, revolutionizing the understanding of efficiency by quantifying useful work.
Author of a 2009 research paper titled 'The Concept of Efficiency: A Historical Analysis,' which inspired the video's discussion on efficiency.
Physicist credited with discovering the equivalence between mechanical work and heat, and whose work helped establish thermodynamics and the unit of energy named after him.
The speaker and author, discussing his upcoming book 'The Deep Life: A Structured Approach' and his philosophy on meaningful living.
A Marxist thinker and theorist who analyzed the concept of efficiency, distinguishing between 'limited efficiency' (mechanical) and 'arbitrary efficiency' (applied to complex systems).
Pioneered the continuous assembly line and used arbitrary efficiency metrics (like Model T cars per labor hour) to improve production, though he also recognized the need for worker welfare and limits.
A computer scientist and innovator who wrote an opinion piece in The New York Times arguing that AI companies are providing tools that make people unhappy rather than useful.
Mentioned in reference to the speaker's consideration of a neon sign in a style similar to Rogan's podcast setup.
Sociologist and criminologist who wrote 'Folk Devils and Moral Panics' in 1972, exploring the concept of moral panic.
More from Cal Newport
View all 328 summaries
67 minPhones Stole Our Lives. We Investigated How to Reclaim Them.
59 minThe Normie’s Guide to Thinking About Tech
67 minWe Had Real People Test Popular Productivity Strategies
79 min“Rogue” AI Is Not Inevitable. It’s a Choice.
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free