All the world's a stage, and all the humans and machines merely players
Is chasing order in an increasingly tumultuous world a Sisyphean or Herculean task?
I used Open Paper to help me parse through the papers and technical reports I referenced in writing this blog post. Thanks to Saba, all my readers can use code VALSTECHBLOG for 50% off the Researcher Plan. Thanks to Ashi (check out her blog) for connecting us!
“The tragedy of the AI commons”
In 2021, computer scientists and philosophers Travis LaCroix and Aydin Mohseni published The Tragedy of the AI Commons. They described how the World Economic Forum had “identified almost three hundred separate efforts to develop ethical principles for AI”, but in large part because there was no way to enforce these guidelines, studies “suggested that such declarations have little to no practical effect”. Referencing ecologist Garrett Hardin’s 1968 formulation of the tragedy of the commons, they identified a core social dilemma (“no one has an individual incentive to cooperate, although mutual cooperation would lead to the best outcome for all those involved”), developed a stochastic evolutionary game theoretic model to simulate models of cooperation at research population scale, and concluded with a set of morals, including “Lowering the cost of cooperation increases the likelihood of cooperative success” and “Small, decentralised groups may benefit sustained cooperation for responsible AI research”. Importantly,
It would also be a mistake to interpret our results and arguments as a dismissal of current proposals or guidelines for the safe and ethical development of AI. To the contrary, we believe that such efforts are laudable and necessary; but to give ourselves reasonable odds for success, we must appraise ourselves of an understanding of the basic dynamics of such coordination problems
Earlier this year, Sahil Bloom published “AI and the Tragedy of the Commons” to his newsletter, providing an industry and life optimization angle to what LaCroix and Mohseni were approaching as a policy and collaboration problem. He outlined “possible (and scary) implications” of the messy socioeconomic incentives currently defining AI development and integration into work and life, most notably a potential future where “People are more efficient and productive, but lose a sense of texture, meaning, or purpose in the work… [having] unwittingly optimized the life out of their life.”
While LaCroix and Mohseni took a theoretical, statistical approach to modeling populations and concluded that costs to collaboration needed to be lowered to offset opposing incentives, Bloom took almost exactly an opposite approach referencing real survey data and anecdotal evidence and landed at the same conclusion:
It’s about how we make it rational for each actor to act with the longer-term future of our shared commons in mind… how we make it rational to work towards a future of human flourishing… In my view, it’s the only question that matters for our future: How can we lower the cost of long-term thinking?
A second-wave creative class
In 2022, Wessie du Toit wrote “The Rise and Fall of the Creative Class”:
By and large… the post-crash [post-2008] world starved and scattered the creative professions, squeezing budgets and forcing artists into the grim ‘independence’ of self-promotion on digital platforms.
One result of this was a general revulsion at capitalism, which partly explains why artisan ideals and environmentalism became so popular in creative circles. But despite this skepticism, and even as career prospects withered, the creative lifestyle maintained its appeal. In fact, the 2010s saw it taking off like never before…
Intellectual property really is the basis for growth and high incomes in today’s economy… the valuable intellectual property is increasingly concentrated in the tech sector. It is largely because IT and software are included that people can still claim the creative industries are an exciting area of job creation…
The tech world is, of course, a very creative place, but it represents a different paradigm of creativity to the arts and media vocations we inherited from the late-20th century. We are living in a time when this new creativity is rapidly eclipsing the old, as reflected by the drop in arts and humanities students, especially in the US and UK, at the expense of STEM subjects.
du Toit’s Rise and Fall came 20 years after Richard Florida’s Rise of the Creative Class. Florida wrote: “The key to economic growth lies not just in the ability to attract the creative class, but to translate that underlying advantage into creative economic outcomes in the form of new ideas, new high-tech businesses and regional growth”. In the years following, many critiqued Florida’s Creativity Index for economic growth (which factored in the creative class share of the workforce, innovation measured as patents per capita, the Milken Tech Pole Index as a High-Tech Index, and a “Gay Index, a reasonable proxy for an area’s openness to different kinds of people and ideas”) as exacerbating existing urban inequality and accelerating gentrification. Frank Bures, in the The Fall of the Creative Class:
[Richard] Florida’s idea was a nice one: Young, innovative people move to places that are open and hip and tolerant. They, in turn, generate economic innovation. I loved this idea because, as a freelance writer, it made me important. I was poor, but somehow I made everyone else rich! It seemed to make perfect sense… Soon after we arrived [in Madison], however, I was sitting at my desk wondering where all these creative, self-employed bohemians might be…
In a four-year, $6 million study of thirteen cities across Europe called “Accommodating Creative Knowledge,” that was published in 2011, researchers found one of Florida’s central ideas — the migration of creative workers to places that are tolerant, open and diverse — was simply not happening.
“They move to places where they can find jobs,” wrote author Sako Musterd…
“Even as an arts advocate,” said Mel Gray, “I want to do it for the right reasons.” The right reason, we can now say, is that these things are good in themselves. They have intrinsic value. They make the place we live more interesting, livelier, healthier and more humane… They do not make it more profitable.
Bures managed to capture in 2012 a sentiment so timeless it could just as easily have been written today in 2025:
…this was Florida’s true genius: He took our anxiety about place and turned it into a product. He found a way to capitalize on our nagging sense that there is always somewhere out there more creative, more fun, more diverse, more gay, and just plain better than the one where we happen to be.
But I’ve been down that road, and I know where it goes. I know that it leads both everywhere and nowhere. I know you could go down it forever and never quite arrive. And I know now that it may be wiser to try to create the place you want to live, rather than to keep trying to find it.
If the economic definition of the creative class was bolstered in the 2010s by tech work, in the 2020s we’re seeing the start of a massive shift in both what is viewed as creative work and how much time we’re spending on “manual / automated” vs. “creative” tasks. Sahil Bloom fears we’ll automate away so much of our work and our lives that we’ll lose the parts of ourselves that make us appreciate being human; Frank Bures warned us more than ten years ago as a creative in the economic transition following the Great Recession that we’d need to grapple with the dichotomy between “intrinsic value” and realized economic value, and actively mold the world we desire from scraps.
The reality is that the sociologists and economists defining and studying the “creative class” and economic ramifications of it are themselves members of a highly privileged, small socioeconomic subset of the world. Just as LaCroix and Mohseni qualified their study as not a critique of ethical AI guidelines, rather a call to action in areas where implementation was lacking, I note that socioeconomic studies of how AI is changing the workforce are not fruitless, but need to be coordinated at a global level and acted upon to shape real outcomes.
The latest Anthropic Economic Index report raises far more questions than it answers
Anthropic recently published updates to its Economic Index (N.B. Still no Anthropic Environmental Index which I was hoping for back in March; guess we’ll have to survive on Microsoft Environmental Sustainability Reports until then), observing that AI (Claude) usage “strongly correlates with income across countries”, “regional usage patterns reflect distinctive features of the local economy, [e.g.] elevated use for IT in California, for financial services in Florida, and for document editing and career assistance in DC”.
If the productivity gains are larger for high-adoption economies, current usage patterns suggest that the benefits of AI may concentrate in already-rich regions—possibly increasing global economic inequality and reversing growth convergence seen in recent decades…
This concentration in advanced economies with limited population sizes reflects their established patterns as technology pioneers. For example, both Israel and Singapore rank highly in the Global Innovation Index—a measure of how innovative different economies across the globe are—suggesting that general investments in information technology position economies well for rapid adoption of frontier AI. Overall, these economies can leverage their educated workforces, robust digital infrastructure, and innovation-friendly policies to create fertile conditions for AI…
The geographic patterns of AI adoption—where it is used, for which tasks, and how—suggest that in order to realize the potential of AI to benefit people across the globe, policymakers need to pay attention to local concentration of AI use and adoption, and address the risk of deepening digital divides.
The report doesn’t propose any meaningful policy suggestion or real-world advice beyond highlighting that regionally different AI usage patterns may become a global economic problem and that policymakers should pay attention.
Researchers affiliated with the National Bureau of Economic Research have just published a working research agenda for “the economics of transformative AI”:
AI capabilities have improved radically in recent years. Our institutions, organizations, skills, and economic models are struggling to keep pace. In this growing gap lie the greatest risks of the coming decade, as well as the greatest opportunities. We urgently need to improve our understanding of AI’s economic implications
They start tackling this problem by proposing precise questions to be answered: “we must define the fundamental questions, laying the groundwork for a productive research agenda… A question well-posed is half answered”:
As we potentially transition to a post-work society, economics will play a crucial role in shaping new institutions to support human flourishing. What frameworks can help us analyze what provides meaning when TAI (transformative AI) can handle most economic tasks? How might the distribution of meaning-generating activities be optimized for social welfare? Can TAI itself help create or facilitate new sources of meaning and fulfillment? How can we equitably distribute the benefits of TAI, not just in terms of material wealth, but also in terms of access to fulfilling activities and meaning?
They also mention the research use of LLM-based agents to simulate large-scale economies, first to drive “new insights into the economy and how AI might reshape labor, consumer behavior, industrial growth, and unforeseen economic bottlenecks”, and then more widely integrated scenario planning “[to bridge] technical, economic research, and policy communities, as well as the broader public”.
This research agenda is a promising and more tangible attempt to focus our understanding and actions around building a future in which humans and machines coexist safely and productively. I appreciate that it formulates many of the questions I had reading the Anthropic Economic Index report into real research areas. I have faith that many of our best and brightest minds are hard at work implementing these ideas, but as the authors repeatedly call out, there is an incredibly urgent unmet need:
While significant resources are being allocated to the technical development of TAI, there has been comparatively little investment in understanding its economic implications and in research to inform policy on how to steer TAI in desired directions and prepare for the impacts.
Economies of taste over scale
I previously wrote about recommendations and curation in a rapidly compressing, increasingly online-first attention economy. Another marker of cultural and regional shifts in consumption we are already observing is the proliferation of recreational “AI slop” (professionally, HBS terms this “workslop”). We’ll start to see a real divergence in the coming decade between consumers (and corporations and governments, other socioeconomic actors) that value and prioritize high-quality products and media and those that begin to accept lower-cost, lower-quality, more automated and mass-produced forms of entertainment and software. We’ve seen this before: the invention and optimization of assembly lines and manufacturing plants to mass produce things like cars, furniture, and mobile homes, turning years, months of human labor into single-digit days; the development of factory farms, pesticides, preservatives, freezer and microwave technology to flip food preparation from a full-time life-sustaining endeavor to a seconds-long afterthought. We doubled, tripled our free time with these inventions. Collectively, we were able to converse more, read and write more, make more art, music, film and television.
But now, having automated away the hard work of clothing ourselves, building and keeping our homes lit, warm and safe, and putting food on the table, having enjoyed the fruits of our ingenuity via new forms of media, we’re now trying to automate away the labor that goes into creating that media (Spotify and Netflix both have complicated relationships with AI-generated material, where in the eyes of critics, Netflix already had a quality problem producing originals too quickly; Suno recently turned a raw recording of a jam session into a shockingly good, professional-level mixed and mastered track for my band, but also completely changed our lead singer’s voice).
A master woodworker who had trained at The Krenov School of Fine Woodworking once told me that the craft of carpentry has seen many forms, early on as a royal luxury, later as a common farm and homestead necessity, and then mostly as a nameless, thankless task in modern homes. But because of how the craft has changed, each individual woodworker now holds incredible responsibility to preserve and advance their craft, adapting to the world’s changing tastes while acknowledging the complex history and industrial processes already involved in their work.
If we put down our tools and turn back because the forest we’re in is too wild and unknown, we miss the chance to carve out a better future for ourselves.