Strong vs. Skinny: The AI Trade-Off
AI amplifies intelligence to make companies productive or thinner
For as long as humankind has developed the power to wield tools we have been expanding our leverage, and augmenting our weaknesses. The wheel gave us greater leverage to travel and move goods that were too heavy or cumbersome to carry. Machines enabled us to replicate tasks with precision, and technology has further enabled us to leverage what’s known as “external cognitive mind.” We need not remember or store information locally, because we can access it or recall it external to our own brains. In a great new film about Umberto Eco, the Italian philosopher and polymath, he comically articulates this as “Mineral Memory” (Earth’s way of remembering history through layers of sediment, fossils, etc), “Vegetable Memory” (Papyrus, pen and paper, and generally the written word of humans), “Organic Memory” (what’s kept inside our heads and bodies), and “Digital Memory.”
The Mosaic web browser helped people navigate IP addresses via more intuitive pages as proxies for machine-readable information addresses. In those early days of the Internet I recall how radio advertisements would always conclude with, “find us on the world wide web at www dot whatever dot com.” This also became an anachronism since the ubiquity of Google. We need not even memorize URLs; we’ve obviated the need to remember any digital address because we can access this information via the search bar. There are many examples, but this notion of leverage and external cognitive mind has existed since the beginning of the modern human era. Homeric poets and masters of oratory lamented the arrival of the written word; I think I’d survive the demise of manual data entry on Salesforce or Google login failures.
“AI” is simply the new term for powerful software as a service (SaaS), and it’s rapidly changing the equation around “build versus buy” in software purchase or development.
We herald the debut of AI as different, but it is not. Artificial intelligence might better be represented as IA, or intelligence amplification, and just as with every choice in history, people and companies will have the choice of “Strong vs Skinny.” What I mean by that is if AI is a 10x multiplier of productivity we have the choice to 1) become ten times more efficient or effective at a given task aided by AI, or 2) operate at the same level of output with one-tenth the effort, or one-tenth the labor. We can always trade off for more leisure, but as a decade-long New Yorker, most of us don’t.
We are beginning to see the world move in two directions. Most are moving toward “Strong,” leveraging, for example, AI co-pilots to make developers writing code ten times more productive. These co-pilots are now also applying to lawyers reviewing contracts, or teachers grading homework. And this isn’t even new. Companies like StitchFix have been leveraging co-pilots for the last decade, supplementing human stylists with Netflix-like algorithms that take into account hundreds of inputs to create a short-list for the stylist to be ten-times more efficient and accurate in their work. Coding co-pilots enable higher quality code with less errors in less time. (Check out my conversation on this in 2017 at the Commonwealth Club in SF).
For companies struggling to maintain growth or profitability they may make an inverse choice in certain job functions to go “Skinny,” rather than “Strong.” In other words, rather than 10x productivity, they’re instead choosing to leverage 1/10 the time or labor input to generate nominally similar results with the aid of IA. I’ve heard rumors from Sweden that this is the direction Klarna is taking, cutting 90% of its copywriting workforce to focus only on higher level roles where editors chaperone AI content generation. With a fraction of the staff they can generate comparable output.
Strong vs Skinny is this decade’s AI choice.
Those who choose Skinny will do so out of survival, and because many will choose Strong, then the whole playing field got more competitive. For those who remember Cobb-Douglas Production Functions from macro economics, Output (Y) is a function of Labor (L), Capital (K), and Technology (A). Technology is the co-efficient that multiplies the stock inputs of how much labor you have available, and capital. As technology like AI improves, this increases the multiplier on the fixed inputs of labor and capital, and leads to greater productivity, or greater output, on the same inputs. Moreover, as AI might be supplementing the efficiency of labor and capital directly as well, one could maybe even argue it becomes an additional multiplier on those inputs.
In the short run technology will create short-run dislocations, or mismatches in skills and roles. But in the longer run the argument that AI will simply eliminate jobs is to take the argument that we’re fine with output the way it is. In other words, hold output constant, increase the technology coefficient, and what happens? Well either labor or capital must get smaller, right? If history is any indication (and of course pundits will offer the cliches I need not even mention), society opts for Strong, not Skinny. Since we’ve had wheels, phones, or the Internet we’ve always expanded the pie, never simply enjoyed it as it were. In other words, as technology increases labor doesn’t simply shrink to keep output constant. Output increases.
AI will be no different. It will amplify human labor and capital to generate output like we’ve never seen. The short run dislocations in labor markets are real, and some companies will absolutely opt for Skinny not Strong, as a means of survival. But I’m of the opinion that the long run prognosis is deeply positive. As AI will increase global output and productivity far faster than population growth, per capita GDP will skyrocket and qualities of life will improve with greater consumption, and therefore demand for production. Of course this will not be equally spread, and there will be inequalities and challenges to climate and sustainable growth, but on the whole, amplifying human intelligence will lead to greater output, not less work.
The motto of the Olympics is “Citius, Altius, Fortius” or Swifter, Higher, Stronger. This ought to be the mentality of intelligence amplification, and the opportunity of IA.
As AI democratizes software development more companies will become SaaS companies rather than purchasing SaaS products. The “Build versus Buy” equation of software purchase or development will change meaning that AI will likely be a solidifying, rather than disruptive, technology. It will benefit incumbents. A company like McDonalds won’t need to purchase as much software, because they can build it.
Some companies will go lean, and we’re certain to laud a few unicorns in the coming years that I would call “Thin Startups.” These will be startups that angle to solve specific problems with the enormous leverage of AI. The wealth creation around these thin companies will be enormous, as very few people with very little capital will be able to create immense output. But this isn’t zero sum, or doesn’t take away from other company abilities to boost output and continue to leverage employment to do ten times what they could before. We maintain the notion that companies operate at full labor efficiency and with more technology we simply won’t need as many people. When I speak to nearly every CEO at every company they tell me about the hundreds of things they would do if they only had more time. We are under-optimized, under-utilized, and we’re only beginning to see the productivity boost AI will power.
For more recent thoughts on AI versus IA see my other recent posts, “Humanities: Our Future Superpower” and “AGI as Distributed Intelligence.” And of course “The Fuzzy and the Techie: Why the Liberal Arts Will Rule the Digital World,” my book written in 2016 that highlights why human skills are vital in the era of AI.
Also available via LinkedIn Newsletter.
—
Scott Hartley is Co-Founder and Managing Partner of Everywhere.vc. You can read more from him here.