[T]he transformations that the [I]nternet facilitated took place in a setting where governments recognized the benefits of innovation and encouraged a light-touch regulatory regime. Now they’ll be playing out in the regulation-heavy workplaces and factory floors of the United States. Our “permission-first” regulatory regime in everything from drones to sidewalk delivery robots means that any company interested in robotics will have to spend time and money dealing with lawyers, permits, and endless regulatory approvals. For more established players, agreements with labor unions make automation maximally difficult: dockworkers unions have been fighting against automation for years, helping ensure that American ports are some of the slowest and least efficient in the world. Simply put, we are not set up to win.

In short, Neuberger and Casado argue that if the US fails to lead in robotics, every downstream pillar of our national competitiveness – from scientific progress to industrial output – will weaken. Long-term American prosperity is dependent on embodied AI being deployed in the real world. And soon. Because the hour is late, and the US needs to act much more boldly if we are to seize this pivotal moment.

Towards a National Robotics Industries Act

In previous posts, I noted the value of the US taking the lead in developing a national robotics strategy, as well as the policy pieces that should compose such a strategy. But a high-level policy wish list, while a necessary precondition for safeguarding US geopolitical interests in this technology, won’t be enough to secure US leadership in robotics.

What the US needs to move from permissioned to permissionless is, ironically enough, legislation that tackles all the robotics policy imperatives in one fell swoop: A National Robotics Industries Act (NRIA).

Corridors of Ambiguity vs Regulatory Umbrellas

Why do we need the NRIA?

Well, just as the permissionless nature of the Internet fueled the growth of US primacy in the digital economy, the permissioned nature of the “world of atoms” (h/t Peter Thiel) will increasingly be the barrier to the US leading in the next wave of general-purpose robotics technology. That’s important because robotics, like AI, isn’t just a sector of the economy – it’s the operating system for the next generation of economic development.

In the digital realm, software development and deployment have largely benefited from “corridors of ambiguity” — those areas of the market that have been left relatively unregulated by sector-specific regulations. This uncertainty allowed new markets in unanticipated products and services to emerge without the regulatory frictions that burden more traditionally regulated industries. This approach has allowed the US to become the world’s engine of digital innovation over the past quarter century, in contrast to policy approaches that prioritize “regulatory umbrellas,” like in the EU, which foreclose the emergence of unforeseen possibilities from these corridors of ambiguity.

This longstanding corridors of ambiguity in digital services has directly contributed to the AI boom in the US. As we move towards a world where real-world applications of this technology become necessary to justify future valuations and market returns, however, that ambiguity will increasingly give way to the existing regulatory umbrella in the US. It’s one thing to develop an AI brain within the corridors of ambiguity; it’s another thing to incorporate that brain into a body that operates beneath the regulatory umbrella.

As a result, we need to take on the arduous task of marginal deregulation across some of the most cost-disease-ridden sectors of the economy — those realms that, incidentally, also happen to bear some of the highest regulatory obligations, which in turn create high barriers to technological disruption and market entry.

So where Neuberger and Casado leave off, policy particulars need to pick up — by creating the conditions where robotics companies can deploy faster, scale quicker, and integrate into those cost-disease centers of the broader economy with minimal friction. What follows are some concrete suggestions for what could be included in the NRIA to help expedite our journey to America’s global dominance of the robotics industry.

A National Robotics Industries Act

The NRIA would do for robotics deployment what the Interstate Highway System did for mobility, and what the CHIPS Act is doing for semiconductors — treat automation as critical national infrastructure. What follows are four general provisions that such legislation should consider — each addressing a vital piece of the broader robotics ecosystem where the US currently lags.

1. Incentivizing American Adoption of Robotics

There are many barriers to robotics adoption in the US. Certainly the regulatory angle is present across economic sectors where cost-disease is rampant, but there are also frictions that emanate from both the cultural zeitgeist and high upfront costs of adoption. One way to address the former is to create an environment where firms — especially small- and medium-size enterprises (SMEs) — are incented towards more widescale adoption and integration. As technologies proliferate and are more ever-present in day-to-day life, people become more comfortable with their presence. More people experiencing a technology’s benefits firsthand is the surest way to build trust and confidence in the technology.

The CHIPS Act proved the US can mobilize industrial policy around strategic technology. Robotics deserves a similar framework, though focused less on direct government grants for research and more on deployment incentives. Small and medium-sized enterprises in particular are at a striking disadvantage when it comes to absorbing the upfront capital costs necessary for adopting both traditional industrial robotics and their newly-emergent advanced variants. They could potentially benefit from tax credits and rebates, though if such a program were to be proposed, it ought to require a post-hoc programmatic analysis to assess whether its usage materially impacted adoption of the technology among SMEs.

Expanding capex support — whether through grants and cost-sharing programs, tax and investment credits, or other mechanisms — that targets individual components of the broader supply chain (i.e., sensors, actuators, harmonic drives, gyroscopes, etc.) is one targeted support vehicle to help incentivize adoption. The DOD Office of Strategic Capital (OSC) model is instructive here. The NRIA could copy the OSC structure and house its variant in a relevant agency like, e.g., the Small Business Administration or Department of Commerce while expanding the program’s definition of “critical technologies” to cover a broader subset of the robotics components.

The NRIA might also include an accelerated depreciation timetable for robotics hardware, with higher immediate cost deductions and phase-outs. The US tax code provides for general tax benefits under its Section 179 provision governing depreciation, but no robotics- or automation-specific depreciation benefits. Such an incentive, paired with targeted grants, could help spur wider-scale adoption of newer robotics systems among smaller firms.

2. Rightsizing Regulations in the “World of Atoms”

When it comes to more advanced humanoid robots, a big barrier to broader adoption is a lack of clear safety standards that creates the market certainty necessary for both investment and deployment. To address this concern, the US should create tiered safety certification regimes that acknowledge the difference between, e.g., a collaborative robot arm, a surgical assistant, and an autonomous quadruped in a warehouse. In other words, standards should reflect real-world use-cases based on the risk a multipedal robotic system poses in a given environment. The National Institute of Standards and Technology (NIST) is the obvious candidate for housing such an effort, but we can go further. The US should establish a National Robotics Authority housed within NIST (NIST-NRA), tasked with four primary purposes:

• Mapping Sector-Specific Regulatory Barriers to Robotics Adoption. In concert with the Office of Management and Budget, the Office of Science and Technology Policy, and sector-specific regulators, NIST-NRA should coordinate on producing an annual review of existing regulatory obligations that pose direct barriers to wider adoption of advanced robotics systems throughout the economy. Such a report, informed by consultations with industry and outside stakeholders, would serve as an invaluable guidepost for how Congress and the Administration can identify, and address, existing bottlenecks holding back US leadership in robotics development and adoption.

• Developing Technical Standards and Guidance. NIST is internationally renowned for its commitment to rigorous technical analysis, and serves as an ideal home for developing best practices and guidance for advanced robotics systems. Those best practices can then serve as the bedrock for wider market, and government, adoption of new robotics systems.

• Engaging in International Advocacy. NIST-NRA can serve an important purpose as the defacto federal agency for industry engagement on robotics issues. It can also serve a central coordination function with other US agencies — and foreign counterparts who share an innovation-friendly approach to governance (i.e., Singapore, Japan) — to push for US-led standards in international standards organizations (i.e., IEEE, ISO).

• Building “Grand Challenges” Roadmaps. Export the “grand challenge” model popularized by DARPA (for autonomous vehicles) and other agencies to robotics. The NIST-NRA can develop proposals for technical milestones and, in conjunction with purchase guarantees from other agencies, can help to both de-risk early deployments and pull emerging tech into national missions.

The NIST-NRA can help to plug many different gaps in the existing federal robotics policy landscape, but its most important function would be to serve as the recognized central coordinating function for federal robotics policy. Importantly, housing this function at NIST, which has no formal regulatory enforcement authority, sends a clear message to ecosystem stakeholders that the new agency’s mission is not to regulate robotics, but to “clear the runway and help achieve escape velocity.”

3. Expanding the Robotics Workforce

As discussed in my last post, robotics shouldn’t be viewed as a job killer, but rather the next stage in the ongoing process of economic automation that began back in the late 18th century. Robotics, much like AI, will augment and redeploy human capital, not destroy it. Nevertheless, policy needs to accommodate the negative impacts of displacement effects on the workforce. Restructuring the social safety net is a big topic, however, and likely beyond the purview of the proposed NRIA.

That said, the workforce issue that does need to be addressed in the NRIA is less about retraining programs (the US has many such programs, such as TAA) and more about the need for talent pipeline creation in advance of wider-spread robotics deployment. A lot more robots means a lot more jobs building, maintaining, and repairing those systems — jobs that require technical education and credentialing for which the US is not currently adequately prepared to fill.

To address this gap, the NRIA should include a provision that establishes a nationally recognized robotics credentialing program, likely housed at the Department of Labor. Such a program needs to be broadly accessible and involve prioritizing partnerships with community colleges and apprenticeship programs, not traditional four-year colleges. Such a credentialing program and associated partnerships should be open to funding from existing educational grant streams. The government can enhance this ecosystem by also making GI Bill funding available for such certification programs.

The next generation robotics economy will ultimately be built by blue-collar technologists. The NRIA should help pave the path for them to lead the world in that endeavour.

4. Securing the US Robotics Stack

AI chips dominate attention, but the robotics stack depends on dozens of upstream chokepoints — harmonic drives, MEMS gyroscopes, compact gearboxes, lidar units, edge controllers, and industrial batteries. China is increasingly in control of much of that stack.

To counter this, the US needs to secure its own robotics stack. In addition to its other purposes, the NIST-NRA should be tasked with working to coordinate a whole-of-government strategic response to securing these vital inputs to the robotics market, including by mapping critical components and identify single-point foreign dependencies and coordinating a more formal Actuator & Sensor Industrialization Program through the Departments of Defense, Energy, and Commerce. The government should also consider how it can potentially leverage reshoring credits or purchase guarantees to suppliers that opt to manufacture critical components here in the US.

Equally important to the supply chain, as Neuberger and Casado note in the a16z article, is access to real-world training data:

As with LLMs, training advanced robotics systems requires pretraining data on the scale of the internet, along with reinforcement learning to train generalist policies that can reason across a wide range of distortions in environment, perception, and task. As data from real-world deployment comes online, the country with more robots gains flywheel momentum; more deployment means more high-quality data which underwrites further deployment.

In a previous post, I argued that one mechanism for addressing the growing gap between Chinese and American firms’ access to this key robotics input, is the development of a National Data Escrow Corporation. Such a construct could, if properly structured, incentivize firms to pool their collective proprietary training data resources, creating a strategic US resource that can be leveraged to help close the gap between the Chinese and the US.

Conclusion

Where AI has seen an explosion in investment, robotics still faces the drag of deployment friction: liability law, procurement lags, safety ambiguity, and fragmented supply chains. The NRIA can help turn corridors of ambiguity into deployment corridors, leveraging everything from US standards development leadership and the depth of our capital markets to ensure America dominates the “world of atoms” in the coming decades much as we have dominated the “world of bits” over the last quarter century.

The robotics race will only be won through a deliberately architected strategy that, above all, prioritizes domestic deployment of new, advanced robotics systems. The US has the capital, the innovators, and the national imperative. What it needs now is coordination, cultural inclination, and focus.

If America treats robotics the way it once treated aviation, space, and semiconductors, leadership is fully within reach.

If not, we’ll end up inheriting the future that others build for us.

Our hunter-gatherer ancestors experienced functionally zero economic growth — life was simply a matter of survival, dependent on nomadically tethering oneself to the movement of wild animal herds. Then, about 10,000 years ago, we figured out the basics of agriculture and, later, animal husbandry, which allowed humanity to begin eking out the barest subsistence. This created some amount of consistency and certainty in daily life, which then required developing institutions and systems for managing time-spatial relationships. The result was the emergence of more complex social arrangements, like religion and government.

But after the Agricultural Revolution, growth stalled again and humanity settled into thousands of years of more marginal improvements in economic growth.

Source: GIS Reports

Economic growth in the pre-Industrial era averaged roughly 0.1 percent per year — essentially, stagnation was the norm for most of human history. Once James Watts’ steam engine entered onto the world stage — around 1760 — that number probably quadrupled or quintupled to 0.4-0.5 percent for the next few decades. By unshackling economic growth from the constraints imposed by the upper bound limits of human and animal energy expenditures, the Industrial Revolution gave birth to an entirely new economic reality. By the modern era, 2-3 percent growth has come to represent normal expectations.

Now, in a world of advanced AI, some throw out pretty wild annualized growth projections of what the future heralds — 5, 10, even 20 percent!

But it seems hard to imagine a world in which the AI we’re all using today somehow manages to quadruple current growth numbers. Sure, ChatGPT is a great advancement in the evolution of search, but it’s not clear how you get from that to immensely higher productivity in, say, manufacturing. This has become an increasing fixture of recent AI conversations — that the new models and systems, while good, aren’t exactly heralding the promised revolution of a post-scarcity society – and many companies, more so than individuals, seem to be having a difficult time making sense of how to leverage this new technology effectively.

As I noted in my initial post, the lack of a clear vision for AI applications and use-cases will inevitably hold back the fullest promise of this technology’s potential. That’s one of the many reasons I’m so bullish on robotics. Robotics and automation offer that compelling vision — one that can easily make up the difference between current economic growth figures and projected growth.

Much is being made about forecasts for the total addressable market (TAM) for AI, with estimates running a pretty wide gamut, from a few hundred billion dollars per year just for direct generative AI to nearly $15 trillion, which accounts for the indirect benefits to productivity improvement and increased consumer demand. Any way you slice that range, it’s a lot of money. But the possible TAM for the entire range of robotics, from industrial applications to humanoid variants, could, by at least one estimate, run in the tens of trillions of dollars.

If you imagine a future where robots move from a backend curiosity largely confined to warehouses and assembly lines to a more ubiquitous feature of everyday life, then that number isn’t beyond the realm of possibility. That future isn’t too tough to imagine, especially in a world where one needs to strain to see anything but diminishing marginal returns in digital sectors (after all, advertising optimization can only do the heavy lifting of adding to GDP for so long).

In short, the future low-hanging fruit in productivity gains will almost certainly be found in the physical realm, for three reasons:

1. The decreasing marginal cost curve of robotics;

2. The historic long-tail value of automation; and

3. The inability of human labour to scale effectively in industries that need more workers.

One way to think about robotics is through the lens of cost curves. For decades, actuators, sensors, and semiconductors made robots too expensive for broad deployment outside of highly specialized settings like automotive assembly lines. But that bill is now falling by double digits annually, and the average selling price of humanoid robots is expected to follow suit. This looks a lot like the early trajectory of solar panels or lithium-ion batteries: once the learning effects kick in and production scales, the marginal cost of each additional unit drops fast. At that point, the economics shift from “expensive experiment” to “why wouldn’t you automate this?”

Still, robotics isn’t just another consumer electronics story. Rather, it’s a continuation of the story that has defined the modern economic growth miracle: automation. The development of the steam engine heralded the Industrial Revolution, but that process has only accelerated since the 18th century. From the cotton gin to the train, the last two hundred years of human history has primarily been a story of freeing human labour from the toils of the fields to be directed towards more creative and intellectual pursuits, which in turn has yielded the scientific and technological breakthroughs that have made further advancements and refinements to automation possible. AI and robots are just the latest chapter in this tale. The value of automation, like other technological advancements, compounds over time, and will almost certainly continue to compound long into the future.

Economists like Daron Acemoglu and Pascual Restrepo have emphasized that robots are what they call “automation capital”: they don’t just complement labour like computers do, they often substitute directly for human tasks. That’s why the productivity gains from robots can be enormous — McKinsey estimates automation could add up to $4.5 trillion to global GDP annually by 2030 — but the distributional effects are messy. Some jobs disappear, others are reconfigured, and entirely new roles spring up around maintenance, programming, and system integration. The 20th-century story of mechanization in agriculture, where labour productivity skyrocketed while total employment in the sector plummeted, is an instructive parallel.

At the same time, the “robots will take all the jobs” narrative misses something important. The bigger problem isn’t too much automation, but too little. We’re still in a world where healthcare assistants spend hours on paperwork, logistics firms struggle with chronic labour shortages, and infrastructure projects are constrained by workforce bottlenecks. One of the fundamental problems affecting these sectors is Baumol’s “cost disease,” where labour-intensive industries tend to be more immune to the technology-driven productivity improvements in other sectors. Healthcare, education, and elder care get more expensive over time because they rely on human labour that doesn’t scale. Robotics has the potential to break that dynamic by introducing productivity growth into precisely those stagnant sectors. If humanoid robots can reliably lift patients, monitor vital signs, or handle routine logistics, the economic ripple effects are transformative.

Of course, none of this plays out in a vacuum. The global competition for robotics leadership is already intense. China currently dominates many of the hardware components (e.g, motors, rare earth magnets, actuators) while the US leads in AI software and chip design. This asymmetry matters. Even as the US maintains its dominance in this technology’s “brain” the “body” of robotics is heavily dominated by non-US firms. History shows that countries which lead in the physical manufacturing layer often set the pace for adoption and standardization. If the US wants to avoid being caught flat-footed, it needs to treat robotics as both an economic and strategic priority.

So what should policy do?

To begin, as I noted in a previous post, the US needs a National Robotics Strategy, if only to create a clear vision for coordinating policy questions related to accelerating the development and adoption of robotics, and the industry more generally. Such a strategy should, however, delve deeper into a number of issues of national importance.

Data escrow and sharing frameworks. For advanced robotics to flourish, developers need access to vast amounts of real-world interaction data. Yet competitive pressures, high costs of generation, and liability concerns often keep this data siloed. A national data escrow system would give firms incentives to contribute without fear of losing proprietary edge. By pooling diverse datasets, the US can accelerate training, improve safety, and ensure robots can navigate the unpredictable messiness of human environments. Without such mechanisms, progress risks being bottlenecked by fragmented datasets and mistrust, leaving the US at a disadvantage in the global robotics race.

Lower the barriers to adoption. Just as renewable energy took off once tax credits and accelerated depreciation made projects pencil out, robotics could benefit from similar tools. Small and medium-sized manufacturers — the firms least able to absorb upfront capital costs — should be able to claim credits for investing in robotics systems. Public procurement can also play a role. If the government becomes an early adopter, piloting robots in infrastructure, defense logistics, or elder care facilities, it both creates scale and signals legitimacy.

Invest heavily in the fundamentals. Robotics isn’t only about AI models; it’s about reliable sensors, lightweight materials, efficient batteries, and robust manufacturing processes. Federal R&D funding should target these “unsexy” components, because they are where supply chains are fragile and where global competitors have an edge. Think of it as the CHIPS Act for actuators and power systems.

Don’t neglect the workforce. History is clear that automation without adaptation breeds backlash. Reskilling programs, community college curricula, and apprenticeships focused on robotics maintenance and programming should be front-loaded into any national strategy. The key is not to prevent automation but to make sure its benefits are broadly shared.

Standards and safety frameworks. If robots are going to operate in warehouses, hospitals, and public spaces, interoperability, cybersecurity, and physical safety cannot be left to voluntary compliance. The US will need to harmonize rules so companies aren’t navigating a patchwork of liability regimes. Without trust, adoption stalls; and without adoption, the productivity gains never materialize.

The economic logic of robotics points toward acceleration: falling costs, rising demand, global competition. The policy logic points toward urgency: build institutions, create incentives, fund the basics, support workers, and standardize safety. The upside is enormous. We could be looking at the biggest infusion of productivity growth since electrification. But history also shows that general-purpose technologies don’t diffuse automatically. They spread because societies built the institutions to guide them — from railroads to the Internet.

Robotics deserves the same treatment. The stakes aren’t just industrial efficiency, but who captures the next wave of global economic leadership. If the US wants to lead in the era of embodied AI, the time to act is now.

I alerted my parents to the discovery of the foreign object, leading the hosts and the assembled adults to file into the piano room for a demonstration. One of the adults inserted a floppy disk into the device, prompting the Steinway to burst to life, keys depressing without human assistance and causing music to ring through the house. I was amazed as this ghost in the machine tickled away at the piano’s ivory keys. That amazement then turned to a grim thought: If you don’t need a person to play the piano, then what’s the point of learning to play? I stopped taking lessons shortly thereafter.

Some time later, I learned that auto-playing pianos were not new. Although that particular Steinway used a modern solenoid system controlled by a computer, pneumatic systems had been around for almost a century. (You probably know the type from old Westerns, where paper rolls with hole punches spin around and use air flow to activate physical levers that depress the relevant keys.) A century of automated pianos and yet we still have human pianists, and people intent on learning to play.

Why?

Scarcity and the Value of Mastery

Let’s first consider a few other realms of social and economic life that have been similarly touched by automation.

We now have AI systems that are about as good as humans in source code generation and programming. IKEA makes mass use of robotics (e.g., computer numerical control routers, automated multi-spindle drilling machines, etc.) to aid in the manufacture — from cutting and drilling to sanding and surface finishing — of their many product lines. Generative AI can be used to write whole novels or create high quality images and videos.

As such, one could be forgiven for thinking that programmers, woodworkers, and storytellers are all dead professions. And yet, just as pianists haven’t disappeared over the past century, neither have these folks. Nor do I see a future in which these and other skilled creative careers will be solely the domain of machine intelligence. That’s because skill isn’t just about executing a function — it’s about knowing why you’re doing it, and shaping the outcome in ways machines cannot. The value of a skill lies in its scarcity, just as the value of creativity resides in intention.

Being a good programmer is fundamentally about knowing how to give properly encoded instructions to a computer. It’s a game of memory and contextual linguistic heuristics. There are a great many people in the world who lack the former skillset but who excel in understanding why they’re building a given application or system.

Everyone needs furniture, and IKEA’s automation processes help to provide access to reasonable quality goods for entry-level purchasers at relatively inexpensive prices. Mass production means it doesn’t cost an arm and a leg to furnish your home or apartment. But as people age and make more money, their tolerance for furniture that needs constant replacement diminishes. At a certain point, people’s willingness to pay for quality and craftsmanship outstrips their pecuniary constraints. Authentic Eames Lounge Chairs, produced by Herman Miller, are still largely built using traditional hand craftsmanship (particularly in the leather upholstery and veneer application). The greater the demand for quality and uniqueness, the greater the need for traditionally skilled woodworkers.

Contrary to the hot mess of writing and production that’s come out of major studios in recent years (I’m looking at you, Disney), crafting a good narrative structure actually isn’t an overly complicated process. Joseph Campbell’s framework of the “Hero’s Journey” — the common narrative structure found in myths, legends, and stories across all cultures and all of human history — can be charted formulaically:

• A hero is called to adventure;

• He crosses a threshold and journeys into the unknown;

• While there, he is tested with trials, makes allies, and confronts adversaries;

• Eventually, he faces a climactic confrontation and gains a boon;

• He then returns to the ordinary world with that boon.

It’s arguably even simpler than that. An old quote incorrectly attributed to Leo Tolstoy suggests that “All great literature is one of two stories: a man goes on a journey or a stranger comes to town.” A generative AI system can replicate the basic structure of a story and even generate a detailed series of images, or videos, to chain together a visual representation of the narrative. But it can’t draw on experience or emotion to make the story meaningful. Stories in particular, and art in general, are fundamentally communicative exercises between an artist and an audience — an expression of an inner human voice made manifest in the real world.

The Studio Ghibli Hypothesis

So returning to the first question: why, a century on from the pneumatic auto-playing piano, do we still have pianists?

Because nobody goes to the symphony to see a piano playing itself.

Learning and mastering an instrument requires dedication to cultivating a skillset. Not everyone can play the piano well. That scarcity elicits appreciation because we know it’s an ability that was earned. And the expression of that earned skill is a testament to a uniquely human endeavour — one that I think is perfectly captured in the person of Hayao Miyazaki.

The Studio Ghibli founder is the source of classic films like Princess Mononoke, Spirited Away, and Grave of the Fireflies. For decades, Miyazaki honed his craft, constantly drawing and rendering even the smallest of details in his film by hand. Indeed, Studio Ghibli is one of the only studios in the world where the production process still relies almost entirely on traditional hand-drawn artistry. In an age where animation has increasingly become a digitized, and even semi-automated, work product, Miyazaki’s disciples continue his legacy by imbuing a decidedly human touch in the studio’s films. Even their most recent 2023 Academy Award-winning feature, The Boy and the Heron, was in keeping with this tradition.

Even amidst a landscape of creative surplus, people crave creative expression that stands apart from the norm and creates a more visceral connection between the artist and the audience. Miyazaki’s films are not beloved because they couldn’t be made by a computer — they are beloved because he refuses to let a computer make them. Each frame bears the marks of choices, revisions, and a lifetime of seeing the world through a singular lens. That is what audiences respond to: the evidence of a person on the other side of the work. This is the Studio Ghibli Hypothesis:

As AI and robotics increasingly saturate the creative market with abundant, technically competent creative outputs, works shaped by a distinct human vision — the auteur ethos — will grow in cultural and economic value.

In a landscape of frictionless production, originality rooted in personal perspective, craftsmanship, and intentionality will become a key differentiator, much as Studio Ghibli’s hand-crafted storytelling still stands apart in a sea of mass-produced animation.

Conclusion

Whether programming code, building furniture, or telling a story, only humans are capable of understanding what people, and the world, might value. That’s one of the many inherent limitations of robots and AI: they don’t create in any meaningful sense; they merely draw from existing patterns and regurgitate on demand. In a world of increasing creative abundance, the human auteur and the niche will not disappear. They will become more valuable — and more meaningful — than ever.

And for what it’s worth, although I quit my lessons, I still play the piano every day.

This platform is dedicated to a simple proposition: that the future, for all its uncertainty, is likely to be better than the present.

Having worked in emerging technology policy for more than a decade, my observation is that assertion has consistently bourne out to be true. Despite how far we have come within a single generation, however, I think some of the greatest opportunities for technology-driven socio-economic betterment are still ahead of us. The current cultural zeitgeist suggests AI will be driving us towards that brighter future. That may be true, but only to a point.

For all the commentary on the current and future value proposition of AI, I actually think the realm of physically “embodied AI” — that is, robotics — is where the proverbial rubber will meet the road in showcasing the economic gains from applied AI solutions. Hence, the name of this newsletter: Positronic Policy.

Many years ago, I wrote a piece about the

moments in history that redefine the human experience. Old ways of thinking are upended, the traditions of yesteryear are turned on their heads, and radical new ideas are not just entertained by a select few, but embraced by the masses. These are periods when societies pivot away from the familiar and comfortable and welcome the future. The famed philosopher and sociologist Karl Jaspers described these periods as Axial Ages.

At the time, I was writing about the prospect of a more general “Technological Axial Age” that would reshape the foundations of modern society. Over the past year or so, however, I’ve started believing that this pivot point is more likely to be catalyzed by advanced robotics, not only in industrial manufacturing but across all sectors of economic and social life. This, in turn, is likely to animate a sweeping reexamination of … well, everything. From the physical and spiritual institutions that form the bedrock of modern society to the moral and philosophical frameworks that define what it is to be human, this Axial Age of Robotics will herald profound changes in the future trajectory of human civilization. And I find that to be an exciting prospect!

This is a coming revolution that few people are talking about — least of all in DC policy circles, where I spend much of my time. I’ve become increasingly interested in this space, so I wanted to use this platform as an opportunity to highlight the beneficial impacts of this technology and provide my thoughts on what policies will help accelerate America’s development and adoption of this technology.

My goal throughout these first few posts will be to detail a set of general principles for why robotics matter, what areas of life this field will inevitably impact, and how our society — and each of us individually — will need to adjust to accommodate those changes. Drawing from these commentaries, my aim is to first assemble something approaching a general treatise on the value proposition of robotics. That will be followed by a regular cadence of thoughts and recommendations on effective policies to promote the benefits, while mitigating the costs, of this technology’s integration into American life.

The potential policy implications of robotics are vast. I’ll be exploring a lot of those particulars in more detail a bit further down the line, but for now, here’s a quick list of some of the issues that have immediately grabbed my attention.

• We don’t have enough training data. Just as generative AI systems need large pretraining data sets to learn, embodied AI needs a similarly significant set of real-world data. Unfortunately, where LLM developers have access to essentially the corpus of human knowledge via Internet crawling, robotics manufacturers don’t have an “Internet of real world manufacturing environments” from which to draw. Collecting data from physical interactions is a much more difficult, and expensive, challenge than simply sending an automated algorithm to scrape the online world, and is currently a major limitation in accelerating advancements in this space — especially for smaller startups. Data policy in robotics is, I think, one of the key policy arenas that can benefit from more creative thinking. (One idea I’m interested in toying with a bit here is some type of data escrow holding company — possibly in the form of a public-private partnership — that incentivizes both large and small players to contribute their unique datasets so as to minimize the data fragmentation that currently plagues the industry. That, or some sort of NAIRR-style shared edge computing infrastructure. Government could potentially create positive incentives for sharing data by kicking in publicly-available data from real-world environments and conditioning access to those data sets on participation in this consortium. This is just an early germination of an idea, but one I’ll be exploring in a future piece.)

• America doesn’t make enough robots. Believe it or not, but over the past half century, the United States has largely ceded the lead in robotics manufacturing to other countries — both allies, like South Korea and Japan, and geopolitical rivals (see: China). That, to me, is a bit crazy. There’s no reason America shouldn’t be at the forefront of this innovation ecosystem, the same as AI. I don’t have an easy solution to the many vectors that have likely impacted this decades-long trend, but an easy starting place is probably the need for some type of national robotics strategy. That will only take us so far, of course, but a journey of a thousand miles begins with a single step. And that seems like a pretty easy first step.

• Who’s at fault when the robots go awry? This is the classic product liability question. Maybe existing liability laws are sufficient; maybe some new framework is needed. Certainly, as robots become increasingly autonomous actors in society, there will be different rules for different applications (e.g., an industrial robot zipping along a factory floor is a helluva lot different than an autonomous vehicle driving down a public highway). But a fundamental question with any new technology is who bears the cost when things don’t go as expected, and as more robots start diffusing across hitherto untouched areas of daily life I expect that question will be at the forefront of a lot of policy conversations.

• The robots are going to take the jobs. We (should) all want the productivity gains from robotics, but how do we manage labour displacement and the nature of work in an increasingly automated society? There’s no shortage of commentary about how the current generation of generative AI systems could portend a culling of white collar jobs, but that technology isn’t going to replace your plumber or electrician. But what happens when an embodied AI can? We’re probably a ways away from that scenario, but it’s worth exploring in a future piece.

And these policy issues barely scratch the surface of the many, many questions presented by an Axial Age of robotics.

Next Up …

But before jumping into those policy conversations, I want to ruminate a bit on the nature of art and creativity in an age of generative AI. This ties directly into the job displacement question, though in a more philosophical light, and how AI does not, contrary to many hot takes, portend the end of storytelling and the artistic endeavour, and my own personal take on what I call the “Studio Ghibli Hypothesis.”

Stay tuned!

Share