From AI Efficiency Gains to Full Replacement: Dalio's Warning, Anthropic's Brakes, and Unmanned Factories

> TL;DR: AI currently delivers 10–100x efficiency gains, but Ray Dalio warns the replacement rate will soon approach 100%. Anthropic is doing three things to prevent global upheaval: AI-assisted manufacturing reshoring, strengthening human connection, and blocking technology distillation. Meanwhile, China is quietly shifting from a "labor-cost advantage" to an "AI-unmanned advantage" — a change the author argues is ten thousand times more consequential than oil prices.

1. Dalio's Warning: From 10x Efficiency to 100% Replacement

Bridgewater founder Ray Dalio made two pointed judgments in a recent interview:

First, AI is currently at the "efficiency boost" stage, but replacement will approach 100%.

> "AI is temporarily just improving human efficiency, like 10x, 100x. But soon, the replacement rate may approach 100% — people completely replaced."

In his 2025 book *How Countries Go Broke*, Dalio compared the current moment to the Industrial Revolution. But his tone has shifted — he previously emphasized "tooling" (humans using AI for decisions), now he discusses "replacement" (humans no longer needed).

Second, nobody knows how long the "window period" lasts.

> "We all view the problem similarly; we just don't know how long the window for the first step will be — maybe two or three years, maybe five or six."

This uncertainty is itself the biggest risk: everyone is planning on "there's still time," but the technology curve may be far steeper than expected.

2. Anthropic's "Brakes": Three Strategies to Prevent Global Upheaval

Anthropic — an AI company valued at nearly $1 trillion — is reportedly working to prevent AI-driven mass-unemployment unrest:

Strategy 1: AI-assisted manufacturing reshoring

> "Manufacturing still needs people for now, replaced more slowly than white-collar work."

The logic: make blue-collar manufacturing a "buffer zone" that absorbs workers displaced from white-collar jobs.

Strategy 2: Strengthening human connection

> "Human connection remains something AI cannot replace."

Trust, emotion, and community belonging are hard for AI to replicate — but can a "connection economy" sustain itself if 90% of productive work is automated? That is the biggest unknown.

Strategy 3: Blocking frontier AI acquisition via distillation

This is the most controversial strategy: maintaining the leader's advantage through export controls, open-source restrictions, and preventing distillation. Yet Dalio himself doubts the effectiveness:

> "IP protection basically doesn't work. Anything publicly used can almost immediately be copied. Most good ideas, if publicly shown and popular, get copied within about six months."

The contradiction — Dalio says blocks don't work while Anthropic pushes them — may be resolved by strategy: even a 6-month delay is valuable if the goal is slowing replacement to give society time to adapt.

3. China's "Silent Revolution": From Labor Cost to AI-Unmanned Advantage

Luo Zhenyu's recent field visits found:

• Energy/chemicals: chemical plants with almost no visible staff, fully automated processes
• Retail: AI shopping assistants, automated warehousing, unmanned delivery as standard
• Manufacturing: a humanoid robot training school with 120 robots in real-world automotive/heavy machinery training

Why does this hit other countries harder?

| Dimension | China | Western countries | |------|------|---------| | Labor market | Flexible, quick layoffs possible | Constrained by electoral politics | | Welfare system | Still developing, high tolerance for transition | Entrenched; mass unemployment triggers political crisis | | AI deployment speed | Government push + corporate execution | Regulatory review + public resistance | | Social acceptance | Pragmatic, efficiency-first | Value conflicts around "AI replacing people" |

The author argues China could gain an overwhelming manufacturing cost advantage within 3–5 years via "unmanned factories + AI management" — not because others don't want to follow, but because they can't given electoral and welfare constraints.

4. Why This Matters "Ten Thousand Times More Than Oil Prices"

Oil prices affect inflation and growth. AI replacement affects how human society exists — when work is redefined, income redistributed, and human value reassessed:

1. The future global order — whoever keeps employment stable keeps society stable 2. What kind of world people live in — "AI lifting humans" (optimistic) vs. "AI replacing humans" (pessimistic) 3. A new definition of national competitiveness — from resource endowments to AI transformation speed

5. Three Unanswered Questions

1. How long can China's "unmanned advantage" last? Other countries face a cruel choice: rapid follow-up (huge social cost), trade protection (losing competitiveness), or accepting dependence.

2. Can Anthropic's brakes actually stop anything? In June 2026, Anthropic called for a global AI development pause, citing "recursive self-improvement" risk:

> "Over 80% of code merged into the codebase is written by Claude; engineers deliver roughly 8x the code per quarter compared to before 2025."

But if AI can recursively self-improve, technology blocks can slow diffusion — not stop the technology's own evolution.

3. Can "human connection" really support an economy? If AI handles 90%+ of production, what standard of living can the remaining 10% sustain?

6. What Can We Do?

For policymakers:

• Launch research on how society adapts to AI — not the technology itself
• Re-examine welfare systems; is UBI becoming a necessity rather than an ideal?
• Govern the *pace* of technology diffusion rather than trying to stop it

For companies:

• Complete AI transformation within the window period
• Redesign organizations: "humans execute" → "humans decide / AI executes" → "AI decides / humans handle exceptions"
• Invest in "human connection" — future differentiation may lie in the "human touch"

For individuals:

• Embrace the "purple-collar" concept (Luo Zhenyu and Prof. Li Yuhui, Renmin University): cross-domain experts who understand business logic and use AI tools for complex on-site problems — projected to exceed 31 million jobs in China by 2035
• Build AI-resistant capabilities: judgment, empathy, creativity, connection — not just coding
• Stay mobile; don't lock yourself into a quickly-automatable niche

7. An Anxiety About Time

If the window is 2–3 years, many people will be replaced before they react. If it's 5–6 years, there's still time to adapt. But exponential growth means by the time you see it clearly, it may be too late — replacement follows an S-curve: slow at first, then suddenly steep.

Anthropic's 80% AI-written code and 8x efficiency gains — this isn't "the future," it's "now."

References

• Dalio, Ray. *How Countries Go Broke: The Big Cycle* (2025)
• Dalio, Ray. Recent interview (June 2026)
• Anthropic. "Recursive Self-Improvement" Blog Post (June 5, 2026)
• Luo Zhenyu. 2026 New Year's speech "Friends of Time"
• Li Yuhui, Renmin University of China. "Purple-collar" concept (*The Book of Predictions: The World in 1000 Days*)
• Fortune. "Anthropic warns AI could soon build itself without human involvement" (June 5, 2026)
• SiliconANGLE. "Anthropic calls for global pause in AI development" (June 6, 2026)