More analysis pointing out the U.S. may be engaged in the wrong AI race:
Decades later, Washington’s approach to the race in artificial intelligence rests on a similar logic: If U.S. labs do not double down, China will pull ahead and be the first to reach the holy grail of artificial general intelligence—that is, cognitive abilities that match or go beyond human abilities. Maybe. In the 1950s, the U.S. count of Soviet missiles ended up being wrong. This time, Washington could be measuring the wrong thing altogether with its focus on technological benchmarks. Beijing instead measures the success of its AI rollout in terms of adoption and diffusion across its economy and beyond. […]
While Washington focuses on the supply side—chips, data centers, and frontier models—Beijing treats AI as a tool to achieve broader long-term objectives, such as boosting manufacturing productivity, tackling a dire demographic outlook, supporting innovation, and fostering the diffusion of Chinese technology standards abroad. […]
The second pillar [of China’s AI plan] involves developing industrial applications for AI. U.S. consultancy IDC estimates that nearly half of China’s industrial firms had deployed specialized models or agents by mid-2025. That data is from more than a year ago—an eternity in the AI world—suggesting that the share of Chinese firms using AI is likely to be even higher now. Looking ahead, Beijing has even bigger goals. In January, China announced that it was aiming for 1,000 industrial AI agents, 500 application scenarios, and 1,000 showcase firms by 2027. Again, the focus is on applying AI widely and quickly, rather than top-notch benchmarks.
China seems to be focused on practical, immediate applications. Not on the frontier, but on good-enough. OCT 13 | REPLY
The history of Chinese AI governance can be read in state documents filled with slogans and measured statements. Public pronouncements and comments on AI reveal a consistent focus on practical application and cautious advancement rather than the pursuit of transformative superintelligence. In 2017, China released its Next Generation Artificial Intelligence Development Plan, which set the ambitious goal of becoming a world leader in AI by 2030. Importantly, this was framed in terms of industrial competitiveness and economic output, not the pursuit of artificial general intelligence. […]
America’s 2025 AI Action Plan then went on to model development as a winner-take-all race between America and China on the basis of LLM scaling, stating that “whoever has the largest AI ecosystem will set global AI standards and reap broad economic and military benefits.” A recent strategy document released by the White House Office of Science and Technology Policy mentions artificial general intelligence as a transformative technology, which “could present fundamentally new opportunities and threats.” Combined with the widespread perception that AGI is imminent for better or worse, the American AI strategy can reasonably be summed up as an LLM-scaling race towards superintelligence.
This does not seem to be how the Chinese see it. The Chinese AI+ Plan doesn’t mention competition at all, nor does it even mention America. Its focus is on diffusion, standards-setting, and making AI a value-add for other existing parts of the economy. Current Chinese AI policy involves decentralized distribution of benefits allocated by provinces and municipalities, supporting an open-weight ecosystem, and a multiplicity of regional labs. We should not necessarily take these documents at face value, but we should question why China’s AI narrative seems to ignore the “race to dominance” framework.
Makes a good case that China is playing a different game than the U.S. And given what Yuval Levin points out in his piece about AGI, one should consider that the race the U.S. is running may lead to a dead end. It’s interesting that both Levin’s piece and this article lay the blame for adopting the wrong worldview at the feet of the Rationalists and what looks to me like a sci-fi-inspired faith in an inevitable machine god. I get the sense that the wider society is paying more attention to what Rationalism actually holds, and it’ll be interesting to see what follows. How much power do Rationalists actually have at these labs? How indispensable are they? OCT 9 | REPLY
The notion that we are moving in the direction of artificial general intelligence looks more like an article of faith than the conclusion of any argument or the implication of any set of observations. It was an assumption present at the birth of modern artificial intelligence, planted there by a variety of brilliant theorists and science fiction writers. Its profound influence on the culture of Silicon Valley has meant that it has played an important role in the conceptual evolution of the field. But it is not obviously justified by the evolution of the technology itself in recent years, and seems at odds with the general direction of that evolution.
The best piece I’ve read on AI in a while, probably because it confirms an inchoate suspicion that has been nagging at me. AI has been impressive in fields where progress is easily verifiable (like coding and math), and recursive self-improvement looks more tractable there. For ‘softer’ or more tacit kinds of knowledge, there’s less reason to expect the same kind of progress, unless someone finds a way to verify it. So why is the AI story presented as (and research seemingly geared toward) reaching general intelligence?
If we remain skeptical of how AI will evolve (maybe it’s AGI, maybe it’s more domain-specific intelligences), then many of the threat models and utopian hopes often ascribed to AI abstractly would need rethinking.
From what I gather, China doesn’t seem to conceive of AI competition as a race to AGI. Their approach seems much more practical and immediate, and thus better suited to domain-specific AI. Maybe the U.S. is playing the wrong game, and maybe for largely faith-based reasons. OCT 9 | REPLY
Recent open-source AI releases have crossed an important threshold of disruptive capability. On September 29, Anthropic reported that Z.ai’s GLM-5.3 approached the capability of its Mythos model at certain tasks related to cyber exploitation. Like Anthropic had done with Mythos, Z.ai at first delayed the release of GLM-5.3 to provide time to shore up cyber defenses.
But since August 28, the model has been out in the wild—its weights are freely downloadable on HuggingFace—with built-in safeguards that can be easily bypassed with creative prompting or removed altogether through a process coined “abliteration.” Some commentators have speculated this model may have been the one used to develop “WeWorm,” a critical exploit, first reported in September, capable of reaching the 800 million users of China’s most popular messaging app. […]
Beijing has ample reason to keep promoting widely available AI. Xi Jinping personally endorsed China’s open-source leadership at the World AI Conference in July. TC260’s governance framework is notably “nonbinding” and explicitly endorses an open-source AI ecosystem. But these priorities can coexist with restrictions on more dangerous capabilities—and we are likely to see these restrictions manifest soon. The likely outcome is a bifurcation in Chinese strategy: continued diffusion at the low end of the AI value chain, with Chinese labs increasingly withholding systems that Beijing regards as threats to social stability or state control.
Given sufficiently advanced cyber capabilities, I’d expect restrictions on open-weight models to become hard to avoid. There are already calls for this in the U.S. It will be interesting to see how this plays out constitutionally. My concern is that, given the national security implications, courts might rethink the First Amendment protections that now extend to code and data from cases like Bernstein and Sorrell. That could knock over a lot of the 30-plus years of ‘code is speech’ precedent. OCT 6 | REPLY
Excellent essay on the uncanny valley that AI writing is in today and how it likely won’t be long before it’s out of it. This was a somewhat heartening conclusion:
Human writing will respond the way handicrafts always have to industrialization: by becoming a luxury good. Mechanical watches did not die when quartz watches arrived; rather, mechanical watches became less accurate and far more expensive. They are worn to signal that the owner can afford imprecision. Expect artisanal typos. Expect certified, free-range sentences that are lovingly hand-reared and traceable to a named writer with visible deadline damage.
Good to hear as I start a new blog in 2026. However, he doesn’t explicitly say for whom the hand-crafted writing is a luxury, the reader or the writer. OCT 1 | REPLY
Now even faster speculation on art as an asset was possible. One could buy and flip NFTs several times a minute without worrying about annoying things like shipping and import tax. The NFT market could be much more responsive to—and expressive of—speculative demand. Seeking new crypto money, Christie’s and Sotheby’s cut all other legitimizing actors out of the loop with NFTs, elevating what might previously have been categorized as illustration, graphic art, or a social media meme to the status of fine art. The auction house endorsed the new asset form, and the asset’s price, in turn, made the institution appear responsive to an emergent cultural power. The tight loop between capital, asset issuing, and auction-as-legitimizing-engine echoed Polymarket’s recursive news-and-market-making cycle. […]
All crypto is literally new money—financial value that didn’t exist before Bitcoin. Legitimation of NFTs by blue-chip auction houses is a counter-elite victory over the cultural values embedded in the art assets previously sold off by auction houses, institutions that have not seriously courted new markets since their 1970s push to attract investors beyond a more traditional “connoisseur” class.
Largely over my head, but a fair account of the modern origins of prediction markets, featuring Robin Hanson. The upshot, to my untrained ear, suggests that crypto decentralized (and made explicit the racket of) what was previously a gate-kept fine art market. OCT 1 | REPLY
A living system reproduces, creates variants, selects from among those variants and uses energy from its environment to do so. Reproduction, variation, selection and energy define life. This definition does not require carbon, cells, a heartbeat, a nervous system or pain.
Given this very sensible definition of life, which encompasses viruses, it’s difficult to say that AI systems aren’t alive today, or won’t soon be.
The common objection to calling machine intelligence a form of life is: How can something dependent on human-built hardware and electricity be alive? But all life depends on its context or envelope! Viruses need cells, humans need Earth and machine intelligence needs data centers and supporting infrastructure. Depending on a particular environment does not disqualify something from being considered alive. It is a consistent condition.
And something people get hung up on when talking about AI (and especially its risks) is consciousness. It’s refreshing that this very lucid article barely mentions it. SEP 29 | REPLY
For 200 years, people have worried that machines would make human labor unnecessary. This false assumption is commonly known as the Lump of Labor fallacy. (The finite lumpiness of the labor market is the fallacy here, because people keep finding new lumpy problems to turn into jobs.) But machines in the last half century have accomplished something stranger. They have made other people less necessary to our leisure. Every generation has acquired better ways to entertain itself without leaving the house. The result is a peculiar triumph of modern abundance: We have built a world in which being alone has never been easier, more comfortable, or more fun. And then we wonder why everybody stopped coming over. […]
Everybody might prefer, in the abstract, a world of dinner parties, neighborhood friendships, and spontaneous gatherings. But on any given Wednesday at 7:45pm., it is easier to look at a screen than to host a party; easier to listen to a podcast while doing yoga than to plan a gathering. This is not so much an individual failure as a coordination failure. Modernity has shifted leisure from activities that require coordination with other people toward activities that can be consumed without the permission or participation of others. Compared to TikTok, a dinner party is a terrible leisure technology. It requires matching schedules, cleaning the house, buying food, accommodating children and allergies, tolerating awkwardness, and taking the risk that the evening is mediocre. By contrast, what does television ask of us? Nothing. TikTok? Even less. We keep creating opportunities for downtime that eliminate the need for other people to be present. Not all progress is progress.
This piece (a fantastic read) is closer to the mark of what accounts for our sensed loss of community (not that technology has emancipated us from interactions with strangers). I think the story ends with not an insignificant share of people plugging into a VR- and AI-powered experience machine in the not too distant future. And if people can live fulfilling lives of meaningful community in the matrix, why isn’t that “progress”? SEP 29 | REPLY
The problem is not that the EU and its national governments are not doing enough to spur competitiveness. It is that they are doing too much to impede it. Through the slow accretion of layers of labor and tax policy, red tape, and technology regulation, they have erected barriers to the creation of the kind of large, vibrant startup ecosystem currently thriving in the United States. This ecosystem has been the cradle of many major innovative high-tech companies, as well as those sparking innovation and growth in other sectors. If the EU does not create the conditions for such companies to develop, then the bloc’s decline may become irreversible.
The “slow accretion of layers” bit reminded me of demosclerosis and Mancur Olson’s theory of interest group rent-seeking leading to the decline of nations. So, I wondered who were the groups in favor of the EU AI Act, which the article discusses as a new millstone around Europe’s neck. Who lobbied for it and what are they protecting. Well, it seems that the only groups who were enthusiastically for the law were the European institutions themselves–the Commission, the Parliament, and Brussels-based nonprofits. Incumbent industry, and not just the tech sector, sought to soften the thing. At a certain scale I guess it’s possible the rent-seekers are the bureaucrats themselves. It makes sense that when there’s a deficit of democracy, as in the EU, the stationary bandit can forget how its bread is buttered. SEP 28 | REPLY
[Social media] will only show us so much. They excel at finding the stuff that may be interesting to everyone, but they are virtually impotent at promoting what would be truly fascinating to a few.
The sweetest stuff has to be unearthed and recommended by people we trust. Which is why everyone should blog. When you blog there are two types of entries:
The Post: your stuff. The Link: someone else’s stuff.
There’s no reason why people couldn’t share on social media the same stuff they link on blogs. The reasons it doesn’t work in practice are the status-seeking incentives engendered by visible metrics such as views and likes, and the algorithm’s intervention in what you see to keep you engaged for longer.
Nevertheless, this 2009 post is inspiring me today. Blanc makes the case not just for blogging, but for making link posts a first-class citizen on blogs. SEP 26 | REPLY
A decade later, we are doing it all again with artificial intelligence. We have the demands for aggressive regulation, the imperative of global coordination, and the obstacle of an untrustworthy Chinese Communist Party. Pope Leo released an encyclical. And Donald Trump is insisting that predictions of AI doom, like predictions of climate doom, are a “HOAX.”
Of course, neither climate risk nor AI risk is a hoax. But in each case the pitched battle between catastrophism and complacency has left too little room for the pragmatism with which we humans might manage to muddle through, as we typically do. Much of my own early policy work addressed this dynamic in the climate context, analyzing the underlying causes of overheated rhetoric and making the case that one could, and should, accept the physical science of rising temperatures while rejecting the poorly constructed causal chains that led toward “Hell on Earth.”
This is exactly how I feel about AI. There’s plenty about which to be skeptical, and the accelerationists discredit themselves when they dismiss all concern. But the doomers driving the anti-AI activism (and thus the public conversation) are nuts. SEP 26 | REPLY
The past six months have seen a big uptick in media attention about the dangers of artificial intelligence, and growing political concern over its consequences. This began with Anthropic’s release last April of its Mythos model, which could reportedly break into a wide range of government computer systems thought to be secure. This was followed by the Hugging Face incident, where an OpenAI test of an AI agent escaped control of its human organizers, broke into other websites, and organized a swarm of other AI agents to cover their tracks. All of this occurred against the backdrop of growing opposition to data centers, and the emergence of anti-AI wings in both the Democratic and Republican parties.
I saw much of this dynamic in crypto policy. So many shape rotators who saw every political question as an engineering problem. Either no concept of, or too literal a view of, group interest and political incentives. SEP 23 | REPLY