✨ Superintelligence soon: How seriously should I take the AI 2040 scenario?

Even if you buy the technological argument, the economic argument isn't great—which makes me wonder about all the worldbuilding and the policy recommendations it helps drives

My fellow pro-growth/progress/abundance Up Wingers in America and around the world:

Based on my half-year of law school and having watched many Dick Wolf-created legal dramas, I’m aware of the falsus in uno, falsus in omnibus doctrine regarding courtroom testimony. If a witness lies about a material fact, the jury can choose to disregard all of the witness’s testimony. False in one thing, false in everything.

Maybe there should be a variant of that legal principle for scenario-building, especially regarding artificial intelligence: absurdum in uno, absurdum in omnibus. If supposedly serious world-building about the Age of AI or the Age of Artificial Superintelligence involves a key premise that’s plainly absurd … sorry, the whole magilla can reasonably be treated as absurd.

That’s kind of where I’m at with AI 2040, the buzzy continuation of AI 2027, an advanced AI scenario exercise published in 2025. In that original report from the AI Futures Project, led by former OpenAI researcher Daniel Kokotajlo, a US-China race toward artificial superintelligence results in either (a) a misaligned ASI that “releases a bioweapon, killing all humans,” or (b) a US company-government committee that controls the ASI, ushering in “a new age . . . that is unimaginably amazing in almost every way.” Ominously, the report also raises the possibility the aforementioned committee “could rely on fully loyal ASIs to maintain their power” in what is basically dictatorial fashion.¹