In a viral essay on X, “Something Big Is Happening,” Matt Shumer writes that the world is living through a moment similar to early Covid for artificial intelligence. The founder and CEO of OthersideAI argues that AI has crossed from useful assistant to general cognitive substitute. What’s more, AI is now helping build better versions of itself. Systems rivaling most human expertise could arrive soon.
While experts know transformative change is coming fast, normies are about to be blindsided. To stick with the pandemic-era metaphor, Tom Hanks is about to get sick.
Between Shumer’s essay and the resignation of Mrinank Sharma — he led Anthropic’s safety team and vague-posted quite the farewell letter warning that “the world is in peril” from “interconnected crises,” while hinting that the company “constantly face[s] pressures to set aside what matters most” even as it chases a $350 billion valuation — well…some people are starting to wig out. Or, more precisely, the folks already super-worried about AI are now super-worrying even harder.
Look, is it possible that AI models will soon indisputably meet various so-called weak AGI definitions, at minimum? Plenty of technologists, not to mention prediction markets, suggest it is. (As a reality check, though, I keep front of mind Google DeepMind CEO Demis Hassabis’s statement that we still need one or two AlphaGo-level technological breakthroughs to reach AGI.)
But rather than technological advances — and I have high confidence generative AI is a powerful general-purpose technology — let’s instead talk about some basic bottlenecks and constraints from the world of economics rather than computer science.
The long road from demo to deployment. The leap from “AI models are impressive, even more than you realize” to “everything changes imminently” requires ignoring how economies actually absorb new technologies. Electrification took decades to redesign factories around. The internet didn’t change retail overnight. AI adoption currently covers fewer than one in five US business establishments. Deploying it across large, regulated, risk-averse institutions demands heavy complementary investment in data infrastructure, process redesign, compliance frameworks, and worker retraining. (Economists term this the productivity J-curve.) Indeed, early-stage spending can actually depress measured output before visible gains arrive.
Richer doesn’t always mean busier. Let’s grant the optimists — and I certainly consider myself pretty darn optimistic — their assumption about fast-advancing AI capability. Output still doesn’t explode on a dime. Richer societies historically choose more leisure — earlier retirements, short workweeks — not more time at the office or factory floor. Economist Dietrich Vollrath has pointed out that higher productivity doesn’t mechanically translate into faster growth if households respond by supplying less labor. Welfare might rise substantially while headline GDP growth stays relatively modest.
The slowest sector sets the speed limit. Even if AI makes some services far cheaper, demand does not expand without limit. Spending shifts toward sectors that resist automation — health care, education, in-person experiences — where output is tied more tightly to human time. (This is the famous “Baumol effect” or “cost disease.”) As wages rise economy-wide, labor-intensive sectors with weak productivity growth claim a larger share of income. The result: Even spectacular AI gains may yield only moderate growth in overall productivity.
The economy’s narrowest pipe. In a system built from many complementary pieces, explains economist Charles Jones, the narrowest pipe determines the flow. AI can accelerate coding, drafting, and research all it wants. But if energy infrastructure, physical capital, regulatory approval, or human decision-making move at ordinary speeds, those become the binding constraints that limit how fast the whole economy can grow.
Economies are adaptive, complex, wonderful systems. They create the physical objects that embody and accumulate complex information — what economist Cesar Hidalgo elegantly calls “crystals of imagination.” And when they change, they adjust through gradual reorganization and reallocation, not through sudden collapse or instant takeoff. I mean, that should be your baseline scenario.
Now, a degree of urgency may be warranted. (Shumer’s advice to embrace the most capable AI tools now and weave them into your daily work seems prudent.) Panic-inducing analogies to early 2020 probably are not.
This piece originally appeared in Pethokoukis’s newsletter “Faster, Please!”
The post What the latest viral AI apocalypse warning gets wrong appeared first on Vox.
