What this covers

Amy Webb and Russ Roberts examine artificial intelligence not as an abstract future technology but as a present force already embedded in everyday systems—from email filters to anti-lock brakes—and shaped by nine concrete companies: six in the US (Google, Microsoft, Amazon, Facebook, IBM, Apple, collectively the G-Mafia) and three in China (Baidu, Alibaba, Tencent, the BAT). Webb argues that the divergence between how these firms approach AI development, combined with the absence of shared global standards, is steering technology toward outcomes that erode human agency and choice. The conversation ranges across how AI systems escape human understanding, why the market fails to punish the Big Nine for breaches of trust, and what geopolitical dangers emerge when China exports surveillance infrastructure globally.

The discussion traces how alien intelligence—self-optimizing systems that grow incomprehensible to their creators—accumulates power through paper cuts): small defaults and nudges that individually seem trivial but collectively remake human life. Webb details China's social credit system and its digital Belt and Road Initiative as concrete examples of AI serving centralized state control, while Webb and Roberts debate whether the solution lies in technological competition, nonprofit conversion via a third way), international regulatory bodies modeled on the IAEA, or cultural change. A key tension surfaces between their positions: Webb emphasizes the need for global guardrails and collaborative governance; Roberts counters that such efforts often fail and that breaking network effects through better design, possibly blockchain-based, remains more realistic. Neither speaker minimizes the no free lunch trade-off at the heart of the problem—any constraint on AI development means surrendering potential miracles alongside potential harms.

Sharpest takeaway

Webb argues that AI is not an abstract future force but is being shaped right now by nine concrete companies in the US and China, and that the absence of shared global norms—combined with China's centrally coordinated strategy and America's profit-driven fragmentation—is steering AI development toward outcomes that erode human agency and choice.

  • AI is developed within specific institutional contexts (the G-Mafia in the US and the BAT in China), not in a vacuum.
  • China's social credit system and Belt and Road digital export show how AI can serve authoritarian control at scale.
  • Small homogenous groups optimizing decisions on everyone's behalf produce cumulative 'paper cut' losses of agency that markets are failing to punish.

The argument · threads4 threads · 24 claims
0.79

Nine companies control AI's future, split between US and China with fundamentally opposed values on privacy and data.

4 pointscentrality 5/5
  • Six of the Big Nine companies are in the United States (Google, Microsoft, Amazon, Facebook, IBM, Apple) and three are in China (Baidu, Alibaba, Tencent).

    “Six of those companies are in the United States--I call them the G-Mafia for short.”

  • Artificial intelligence is overwhelmingly being funded and built by just nine large tech companies—six in the US (Google, Microsoft, Amazon, Facebook, IBM, Apple, called the 'G-Mafia') and three in China (Baidu, Alibaba, Tencent, called the 'BAT')—who build the open-source frameworks, tools, data sets, and deploy AI at scale, making them the de facto architects of humanity's AI future.

    “In reality, there are 9 big tech giants who overwhelmingly are funding the research--building the open-source frameworks, developing the tools and the methodologies, building the data sets, doing the tests, and deploying AI at scale.”

  • The core danger is the absence of a singular set of global guardrails, norms, and standards for AI—no agreed-upon ideas about who and what to optimize for, under what circumstances, or what data sets to use—while the US and China hold fundamentally different views on what aspects of private life should be mined and productized.

    “What concerns me is that we do not have a singular set of guardrails that are global in nature. We don't have norms and standards.”

  • Unlike US tech firms that grew in an antagonistic, competitive, lightly regulated environment with overlapping products, China's BAT companies operate in lockstep with Beijing, which centrally assigned each company an AI focus area (Tencent on health, Baidu on cloud and AI transportation, Alibaba on data), producing a streamlined system without regulatory infighting.

    “It was centrally coordinated. And that's a really, really important thing to keep in mind.”

0.79

China deploys AI surveillance globally and domestically in ways that entrench authoritarian control and could exclude non-compliant actors from future economic participation.

4 pointscentrality 5/5
  • Through a digital version of its Belt and Road Initiative, China is exporting AI surveillance technology to roughly 58 pilot countries—many with fragile social stability and weak free-speech protections—where, absent China's cultural context, the technology becomes a dangerous weapon in the hands of authoritarian regimes.

    “What does it mean for 58 pilot countries to have in their hands a technology capable of mining and refining and learning about all of their citizens, and reporting any infractions on up to authorities?”

  • China is deploying AI-powered recognition technology (facial recognition, gait and posture analysis) to continually mine and refine citizen data, automatically identifying infractions like jaywalking, publicly shaming offenders via digital billboards and social media, and aggregating behavior into a Social Credit Score that affects access to credit, travel, and schooling.

    “cameras that are embedded with smart recognition technology will automatically not just recognize that there's a person in the intersection when there's not supposed to be, but will actually recognize that person by name.”

  • If China's social credit system evolves alongside its global digital infrastructure investments, future trade conflicts could escalate into a scenario where people, businesses, or even whole countries without an adequate Chinese credit score are locked out of participating in the Chinese—and thereby part of the global—economy.

    “what happens if this credit scoring system evolves and China does business with, only with countries that have a high-enough score? We could quite literally get locked out of part of the global economy.”

  • China's social credit system makes cultural sense as a modern extension of a thousands-year-old community-reporting ('tattle-tale') mechanism used to maintain social control over vast, geographically dispersed populations, which is why outcry inside China is far lower than outside it.

    “One distinctive feature of China is a community-reporting mechanism that is sort of embedded into society. And going back many thousands of years”

0.71

AI must be governed as a global public good with binding universal standards rather than left to market forces or national competition.

5 pointscentrality 4/5
  • Preserving the status quo—where the G-Mafia work competitively rather than collaboratively—gives China a strategic advantage that will worsen over time and likely sets US firms up for a problematic regulatory environment, so a fundamentally different way forward is needed.

    “Preserving the status quo gives China a strategic advantage that is going to become a problem for us the further along that we go.”

  • The marketplace fails to punish the Big Nine for breaches of trust not merely because they make life better, but because modern society and large parts of the business world literally cannot function without Amazon, Google, Microsoft, IBM, and Apple, leaving consumers without a real choice to defect.

    “the reason is not because Google, Amazon, Apple, IBM, Facebook, and Microsoft make our lives a little bit better, but rather that our lives don't work without these companies.”

  • The Amazon Basics Alexa-enabled microwave illustrates how seemingly trivial products are actually data-collection tools: by tracking consumption (e.g., popcorn usage) Amazon optimizes subscription replenishment and locks customers into its ecosystem, foreshadowing a future where non-interoperable devices create 'Amazon families' or 'Google families' whose data is heritable and switching is prohibitively costly.

    “this is a way for Amazon to mine and refine my data in order to optimize that popcorn delivery specifically for me.”

  • Rather than regulation, AI should be treated similar to a public good like air, establishing that everyone has a stake and enabling global guardrails through widespread collaboration on specific agreed-upon tenets—such as requiring that a portion of every AI investment be allocated to safety or cleaning training databases—possibly overseen by a global body modeled on the IAEA.

    “one way, I think, that I think that we could think about the future of AI, is to treat it, you know, similar to a public good, the way that we might treat air.”

  • Artificial intelligence is not a future event with a single arrival moment ('no event horizon'); it is already here, powering email and anti-lock brakes, and represents a Third Era of computing (after tabulation and programmable systems) defined as systems that make decisions on our behalf using optimization.

    “The challenge with artificial intelligence is that it's already here. It is not--there's no event horizon. There's no single thing that happens.”

0.65

AI systems increasingly make decisions we do not understand, eroding human agency through invisible defaults and self-training algorithms that escape human comprehension.

4 pointscentrality 4/5
  • Calling these systems 'artificial intelligence' is a misnomer because it wrongly assumes they resemble human thinking, when in fact we don't understand our own brains well enough; a better term is 'alien intelligence,' because as self-propagating systems they grow more alien to us and begin making higher-stakes decisions in ways we don't understand—with the original creating group functioning as a 'God in that system.'

    “When we say 'artificial intelligence,' I think that that's actually a misnomer... What's probably a better term is 'alien intelligence,' not 'artificial intelligence.'”

  • AI systems are no longer static; they continually learn, join millions of other optimizing algorithms, and—as shown by DeepMind's AlphaZero learning multiple games from zero knowledge and creating child AIs—are increasingly capable of training themselves in ways that defy human understanding, meaning we are slowly ceding consequential decision-making to systems we don't comprehend.

    “And the problem is that these systems aren't static. They continue to learn.”

  • AI-driven recommendations and nudges arose as an antidote to the 'tyranny of choice' created by the modern Internet's overwhelming abundance, but they embody decisions made by someone else about what is best for the user, with no one asking what the individual actually wants.

    “You could argue that using AI to make recommendations was simply an antidote to the tyranny of choice which we created for ourselves in the early days.”

  • The loss of agency from AI systems built by small homogenous groups won't arrive as a single catastrophe but as accumulating 'paper cuts'—small unrequested defaults you can't override, like a car that auto-lowers the radio in reverse—that individually seem trivial but collectively transform life into something miserable and unlike before.

    “To me, this is like a paper cut. And, the challenge with paper cuts is that you get one or two, and you don't sort of notice them.”