What that viral “Something big is happening” AI post gets wrong

In a viral essay about X: “Something big is happening“Matt Shumer writes that the world is experiencing a similar moment to early Covid for artificial intelligence. The founder and CEO of OthersideAI argues that AI has gone from being a useful assistant to being a useful assistant general cognitive replacement. In addition, AI is now helping to develop better versions of itself. systems competes with most human expertise could arrive soon.

While experts know that change is happening quickly, ordinary people are about to be caught off guard. To stick with the pandemic-era metaphor: Tom Hanks is about to get sick.

Between Shumer's essay and this resignation by Mrinank Sharma – he led the security team at Anthropic and vaguely posted quite The farewell letter warned that “the world is at risk” from “interrelated crises,” while suggesting that the company was “constantly faced with this.”[s] “The pressure to put aside what’s most important” even as the company seeks a $350 billion valuation — well, some people are starting to back off. Or more accurately, the people who are already very worried about AI are now even more worried.

Look, is it possible that AI models will soon indisputably encounter various so-called? weak AGI definitionsat least? Many technologists, not to mention prediction markets, suggest this. (However, as a reality check, I'll keep it in mind Statement from Demis Hassabis, CEO of Google DeepMind that we need one or two more Technological breakthroughs at AlphaGo level to achieve AGI.)

But instead of talking about technological advances – and I have great confidence that generative AI is a powerful general-purpose technology – let's talk about some fundamental bottlenecks and limitations from the world of business, rather than computer science.

The long road from demo to deployment. Jumping from “AI models are more impressive, even more than you think” to “everything changes imminently” requires ignoring how economies actually absorb new technologies. It took decades for electrification to transform factories. The Internet didn't change retail overnight. AI adoption currently includes less than one in five US companies. Deployment in large, regulated, risk-averse institutions requires significant complementary investments in data infrastructure, process redesign, compliance frameworks and workforce retraining. (Economists refer to this as “ Productivity J-curve.) In fact, early-stage spending can actually lower measured performance before visible gains occur.

Richer doesn't always mean busier. Let’s give the optimists – and I certainly consider myself quite optimistic – their guesses about rapidly advancing AI capability. The output still doesn't explode in the blink of an eye. Richer societies traditionally prefer more leisure time – earlier retirements, short work weeks – rather than more time in the office or factory. The economist Dietrich Vollrath has pointed out that higher productivity does not automatically lead to faster growth if households respond by supplying fewer workers. Welfare could rise significantly while overall GDP growth remains relatively modest.

The slowest sector determines the speed limit. Even if AI makes some services significantly cheaper, demand is not growing indefinitely. Spending is shifting to sectors that resist automation – healthcare, education, personal experiences – where output is more tied to human time. (This is the famous one “Baumol effect” or “cost disease”). As wages rise across the economy, labor-intensive sectors with weak productivity growth take up a larger share of income. The result: Even spectacular AI successes may only lead to moderate growth in overall productivity.

The narrowest pipe in the economy. In a system consisting of many complementary parts, explained According to economist Charles Jones, the narrowest pipe determines the flow. AI can speed up coding, designing, and researching at will. But if energy infrastructure, physical capital, official approvalor human decision making When things move at normal speeds, they become binding constraints that limit the rapid growth of the entire economy.

Economies are adaptable, complex and 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, adaptation occurs through gradual restructuring and reallocation, not sudden collapse or immediate take-off. I mean, that should be your base case.

Now, some degree of urgency may be warranted. (Shumer's advice to take advantage of the most powerful AI tools now and integrate them into daily work seems wise.) Panic-inducing analogies to early 2020 probably aren't.

This article originally appeared in the Pethokoukis newsletter “Faster, please!”