China just proposed a finish line for the AGI race, and it's denominated in dollars, not benchmarks.
The Summary
- MiniMax co-founder Yeyi Yun says AGI should be measured by economic output: 1% of global GDP, not academic tests or chatbot benchmarks
- This reframes the entire AGI conversation from "can it think?" to "can it build?"
- At current global GDP (~$105 trillion), that's a $1.05 trillion threshold — roughly the size of Indonesia's entire economy
The Signal
MiniMax isn't interested in the Turing Test. The Chinese AI lab just proposed a new metric for artificial general intelligence that skips past the philosophy and cuts straight to production: AGI arrives when AI systems can autonomously generate 1% of global economic output.
This is not how the West has been thinking about AGI. OpenAI, Anthropic, and DeepMind have spent years building evaluation frameworks around reasoning, planning, and multi-modal understanding. MiniMax co-founder Yeyi Yun just said: forget the benchmarks, show me the GDP.
"The milestone for AGI will be set by its ability to generate 1% of the global economy."
The math makes this concrete. Global GDP sits around $105 trillion. One percent is $1.05 trillion in annual economic output. That's larger than the entire economies of Saudi Arabia, Turkey, or Switzerland. It's not a sci-fi threshold. It's a business target.
What does AI-generated GDP even mean? Not ad revenue from chatbots. Not efficiency gains from better code completion. Real autonomous production:
- AI agents running entire supply chains without human oversight
- Synthetic media production replacing human creative work at scale
- Autonomous research labs discovering and patenting new materials
- AI-managed funds allocating capital across global markets
MiniMax's framing also reveals a strategic split. Western AI labs are optimizing for safety театrics and incremental product releases. Chinese labs are building for economic displacement. This isn't about beating humans at chess or passing the bar exam. It's about replacing human economic activity wholesale.
The 1% threshold has another implication: it's achievable in the next funding cycle. If you assume AI productivity tools and agents already contribute 0.1-0.2% of global GDP (conservative, given Microsoft's Copilot revenue alone), then a 5-10x scale-up gets you there. That's not a moonshot. That's a roadmap.
The Implication
If AGI is a GDP target, the race is no longer about who builds the smartest model. It's about who captures the most economic surface area. Watch which AI labs are partnering with logistics companies, manufacturing chains, and financial institutions. That's where the real competition is happening.
For anyone building in the agent economy: this validates the thesis that autonomous economic agents are the unlock, not better chatbots. The companies that win Web4 will be the ones that give their AI the keys to production, not just information.