This post was originally published on this site.
—
My fellow pro-growth/progress/abundance Up Wingers in America and around the world:
A scenario to consider: If the US federal government were ever to quasi-nationalize the big artificial-intelligence labs, it might look something like what investor and AI researcher Leopold Aschenbrenner proposed in a viral 2024 analysis. He envisioned a catalyzing moment when, as AI rapidly approaches the inflection point of general or even superintelligence, Washington would fold the labs and leading cloud providers into a government-orchestrated—but voluntary—consortium for national-security reasons.
Exciting stuff! Think of it as a 21st-century version of the Manhattan Project (though the AI-nuclear comparisons typically obfuscate more than they illuminate).
Now, I’m not even sure that scenario is all that likely, even assuming the conditional technological advance.
But far less likely, I think, is what Bernie Sanders just proposed in The New York Times. Even though AI is currently neither (a) disrupting the labor market nor (b) supercharging productivity growth, the socialist US senator from Vermont would have the federal government ASAP seize 50 percent of the stock in the largest AI companies—OpenAI, Anthropic, and xAI among them—through a one-time tax paid in stock.
Hey, no time like the present. I guess.
Washington would also get board seats and voting power at each firm. Sanders sees the returns from this unique sovereign wealth fund—which he likens to Norway’s oil-funded sovereign vehicle—flowing to Americans as direct payments, eventually funding health care, education, and housing.
Understanding his supposed justification is crucial. Sanders:
Since A.I. is built on the collective knowledge of humanity, the wealth it generates must benefit humanity. Not just Mr. Musk, Mr. Altman, Dario Amodei and other moguls whose companies are positioned to dominate the industry. Not just venture capitalists in Silicon Valley or money managers on Wall Street who undoubtedly see A.I. as the next great wealth-extracting machine.
A genuine question: What is it that Sanders thinks the American private sector does, exactly? Does he see Corporate America as adding any value at all from taking an idea and then improving and commercializing it? Not understanding the gap between the lab and the marketplace is a fundamental error on his part, one of several that help explain this terrible idea. As they say on YouTube and TikTok, let me explain:
First, Sanders is confused about how innovation benefits society. Unless there’s a government-run scheme to deliver Americans a direct check, he seemingly dismisses—or enormously downplays—the other ways AI might provide benefits. Maybe also consider, I dunno, cancers diagnosed earlier and treated more effectively, productivity gains that raise incomes, cheaper goods and services, new products and industries, and breakthroughs across the many of fields of science.
Worse, his view of innovation and the role of business recalls a related idea from fellow Down Winger Alexandria Ocasio-Cortez, who has previously argued that because the federal government funds basic research in the US, the public acts as a de facto early-stage investor in the resulting products—drugs, for instance—yet receives no return on that investment since no licensing deals bring money back to the NIH.
Yet the whole point of this wildly successful system of American innovation is to transfer knowledge to private industry without licensing deals or royalties so as to minimize as much as possible friction between public research and private commercialization. Of course, that only makes sense if you think something useful happens during that business commercialization, and I’m not sure AOC or Sanders do.
And even setting aside the tax revenue and economic growth such research generates, a royalty check to Washington isn’t the only—or even the most important—way taxpayers can benefit from federal research. How about better health from a new pancreatic cancer drug that keeps patients alive twice as long?” Likewise, the most important “return” on AI for the American public would not be measured in dividend checks from a sovereign wealth fund.
Second, Sanders is confused about the trade-offs from AI. A skeptic might say the senator presents AI in whatever way fits his argument of the moment. He has repeatedly characterized AI as a possible existential threat to humanity if it becomes supersmart and turns on us. To me, that sounds like a level of intelligence in the ballpark of what would be needed to also generate the economic value Sanders hopes to extract from the AI companies. Risk and reward go together. And given how much Sanders has stressed the risk, shouldn’t he really be pushing “stop AI” rather than “let’s spread the wealth from AI”? A government that owns half of OpenAI’s stock has a powerful financial incentive to keep the company growing, not to slow it down. And, again, AI with the sort of capabilities Sanders frets about would also be capable of producing the upsides he skates over. (Not to mention, he simply assumes superpowerful AI is going to happen and wants to thus act preemptively.)
Third, Sanders is confused about AI being like oil. Here we go again. He wants to seize half of America’s most valuable business sector and build a sovereign wealth fund on top of it—all that using an Age of AI version of the “data is the new oil” metaphor that became a tech-policy cliché in the 2010s to describe how Facebook and Google turned user data into better-targeted advertising. Back then it spawned a whole “pay me for my data,” or “data dividend,” movement.
The analogy was dumb then, and it remains dumb. Unlike oil, the value from data comes from what you do with it, not from owning a scarce pile of it. A fresh dataset is worth little until it’s combined with other data and run through algorithms created through engineering talent and capital. Oil, then, is the product. AI training data is merely an input. Norway taxes crude that sat inert underground before energy producer Equinor arrived. But no working GPT model existed before OpenAI. The accumulated knowledge was necessary but came nowhere close to sufficient. What Sanders wants to extract is the value created after that knowledge was gathered, but he seems to miss the trillions in value that years of engineering and billions in risk-tolerant capital brought into being. All and all, this starts to sound like an idea driven by politicians looking for an excuse to reduce corporate power for ideological reasons.
Since I don’t think Sanders views this proposal as leading to passable legislation anytime soon, I’m not surprised his NYT piece offers little more than a hand wave on implementation and business impact. (“More details — including the specific spending priorities and the mechanics of implementation — will be included in the legislation I unveil in the coming weeks.”)
Nothing on how an ahistorical equity seizure would crater the market value of the very companies he’s counting on to fund his sovereign wealth fund.
Or dry up the venture capital investment that is a critical American competitive advantage over China.
Or drive away key talent to friendlier jurisdictions, undermining those companies the government just invested in.
Or risk government-directed crony capitalism mismanaging these megacompanies.
Or trigger the myriad other possible consequences of an unprecedented act of government intervention in private enterprise.
But why sweat the inconvenient details of reality when you’re making a social-media-ready political argument to prime the pump for future action rather than a serious public-policy one for this or the next Congress?
