Why I'm Not Worried About AI Job Loss (And You Shouldn't Be Either)

The Short Version

• AI is transformative. But transformative does not mean “mass unemployment next year.”
• Comparative advantage matters more than absolute advantage — even if AI is better at everything, humans don't get replaced
• Human bottlenecks (regulations, politics, company culture, resistance to change) slow adoption far more than anyone predicts
• Jevons Paradox: more efficient tools historically create more jobs, not fewer
• GPT-3 is six years old. Not even outsourced customer service has been fully automated. Think about why.
• The real danger isn't AI — it's panic-driven political backlash that could halt progress

“Something Big Is Happening” — But It's Not What You Think

Let's start with what Shumer gets right. AI capabilities are advancing fast. Models like Claude Opus 4.6 and GPT-5.3 Codex can write working software, draft legal memos, and analyze financial data at a level that would have been science fiction five years ago. Nobody serious disputes this.

But Shumer makes a leap that doesn't follow from the evidence. He compares this moment to “February 2020” — the weeks before COVID shut down the world. The implication is clear: just as people who didn't prepare for COVID got blindsided, people who don't prepare for AI job loss will get blindsided too.

This comparison is wrong, and it's wrong in an important way.

The difference matters. A gradual transition gives people time to adapt, retrain, and find new roles. A sudden shock doesn't. Shumer is telling people to prepare for a sudden shock. The evidence points to a gradual shift. These require very different responses.

Comparative Advantage: Why “Better” Doesn't Mean “Replaced”

This is the single most important economic concept that the doomsayers ignore.

Shumer argues that AI can now do most knowledge work better than most humans. Let's grant that premise entirely. It still doesn't mean humans get replaced. Here's why.

In economics, what matters isn't absolute advantage (who's better at a task in isolation) but comparative advantage (who has the lower opportunity cost). Even if AI is better than you at everything, it can't do everything at once. The tasks where AI's advantage is largest get allocated to AI. The tasks where AI's advantage is smallest stay with humans. Both sides benefit from this trade.

This isn't theoretical. We're already seeing it play out. Software engineer job postings have actually increased since tools like Claude Code launched. Why? Because “Human + AI” teams are more productive than AI alone. Companies don't fire their engineers when AI makes them faster — they give them bigger projects.

The person who understands the client, knows the regulatory landscape, navigates internal politics, and applies judgment where the stakes are high — that person just got a powerful new tool. They didn't become obsolete. They became more valuable.

The Human Bottleneck Regime

Here's something the AI doomsday essays never address: the world runs on humans, and humans are slow.

AI capability is not the bottleneck for adoption. The bottleneck is everything else. Look at what actually slows down technology adoption in the real world:

These bottlenecks aren't bugs. They're features of how human civilization works. And as long as they exist — which is to say, as long as humans are running the world — humans remain complementary to AI, not substitutable by it.

Jevons Paradox: Why More Efficient Tools Create More Jobs

In 1865, economist William Stanley Jevons noticed something counterintuitive. England was getting more efficient at using coal. You'd expect coal consumption to go down. Instead, it went up. Way up. Why? Because efficiency made coal-powered activities cheaper, which massively increased demand for them.

This pattern has repeated across 161 years of technological progress, and it applies directly to the AI jobs debate.

Shumer's essay acknowledges that AI makes knowledge work cheaper and faster. But he doesn't follow the logic to its conclusion. If AI makes legal research 10x cheaper, companies don't fire 90% of their legal team. They do 10x more legal research. If AI makes financial modeling 5x faster, firms don't cut 80% of their analysts. They run 5x more models and explore 5x more strategies.

Demand is not fixed. That's the fundamental error in every “AI will take X% of jobs” prediction. They assume the pie stays the same size. It never has. It never will.

Six Years of GPT — Where Are the Layoffs?

GPT-3 launched in June 2020. That was six years ago. GPT-4 launched in March 2023. That was three years ago. Claude, Gemini, Llama, and dozens of other capable models have been widely available for years.

So where are the mass layoffs?

Think about the lowest-hanging fruit. Outsourced customer service — call centers in the Philippines and India handling scripted interactions for American companies. If any job should have been automated by now, it's this one. The models are capable. The cost savings would be immediate. The work is largely routine.

This should surprise the people predicting imminent doom. If we haven't even automated the easiest, most obvious target in six years, what makes anyone think we'll automate lawyers, doctors, and financial analysts in the next two?

The “any day now” predictions have been running for years. At some point, the predictors need to account for why their predictions keep not coming true. The answer is the bottleneck regime. It's not going away.

The Real Danger Isn't AI — It's Panic

Here's what actually worries me.

Essays like Shumer's — well-intentioned, written by smart people — terrify ordinary people. A parent reading “nothing that can be done on a computer is safe” doesn't think about comparative advantage. They think about their kid's future. A mid-career professional reading “50% of entry-level jobs will disappear” doesn't think about Jevons Paradox. They think about their mortgage.

AI is going to help us cure diseases, solve climate challenges, make education accessible, and increase productivity in ways that raise living standards globally. Shutting that down because of overblown job loss predictions would be one of the worst policy mistakes in history.

We need clear-headed analysis, not breathless warnings. We need to help people adapt to a changing world, not convince them that the sky is falling.

Human + AI: How Smart Professionals Are Actually Working

The winning strategy isn't to compete with AI. It's not to hide from it either. It's to pair your irreplaceable human skills — judgment, relationships, domain expertise, creativity — with AI's speed and scale.

This isn't a feel-good platitude. It's what the data actually shows. In study after study, “Human + AI” teams outperform AI working alone. Why? Because humans catch the errors AI misses. Humans know what questions to ask. Humans understand the context that doesn't exist in any dataset. Humans have the relationships that turn analysis into action.

In each case, the human didn't become obsolete. They became more capable. The AI handled the grunt work. The human handled the work that matters.

This is why tools built around the “Human + AI” philosophy matter. Not AI that tries to replace your workflow, but AI that fits inside it. AI that gives you a research analyst for every client, lets you query your own documents, and works system-wide across your Mac — in your email, your Word documents, your Slack, your Notion — without you leaving the app you're already in.

That's the approach behind Elephas. It's designed for professionals who want AI to amplify their expertise, not replace it. Your documents stay on your device. Your client knowledge stays private. You stay in control. The AI serves you — not the other way around.

That's not a pitch. It's a philosophy. And it's the right one for this moment: use AI to become better at what you do, rather than worrying that AI is going to do it without you.

The Ordinary Person Will Be Fine

Will AI change jobs? Yes. Will some roles evolve significantly? Absolutely. Will there be disruption and adjustment? Of course. Every major technology in history has caused that.

But an avalanche? Mass unemployment? A “February 2020 moment” for the job market? No. The economics don't support it. The history doesn't support it. And six years of real-world evidence don't support it.

References

• Matt Shumer, “Something Big Is Happening” (2026)
• Jevons Paradox — Wikipedia
• Comparative Advantage — Wikipedia
• CNBC: Dario Amodei warns AI could displace half of entry-level jobs
• Statista: Global software developer population growth