Leadership1 publisher2 min readPublished
A Noahpinion roundup splits the opposition to an AI slowdown into five separate interests
The post argues the coalition against slowing AI down is losing the public argument, and it rests that on bipartisan majorities it reads off Nate Silver's summary of the polling. The five constituencies it names each want something different.
The Board Room · Leadership desk
What happened
- A Noahpinion roundup says AI companies and researchers generally believe the technology they are building could do great harm, perhaps even end the human race, if it is not developed more slowly.
- It counts five constituencies opposing a slowdown: the Trump administration, investors, libertarians and techno-optimists, China hawks, and progressives who spent recent years arguing AI does not work.
- The post argues that coalition is losing in the court of public opinion even though it has the president's support for now.
- Inside the progressive movement it reports a fight between an "AI is dangerous" camp led by Bernie Sanders and an "AI is fake" camp that includes his former communications director, David Sirota.
- Barack Obama is telling Democrats to make AI safety one of their tentpole issues, according to the roundup.
Compiled by The Board RoomSomething wrong?How this is made
Why it matters
- constraint The five groups each want a different thing, so one concession quiets one of them: a liability safe harbour aimed at investors leaves the China-race argument and the objection on principle untouched. A compliance plan has five counterparties.
- exposure If the bipartisan majorities described hold, deployment choices made this quarter get argued in public on safety grounds, and the company that shipped the feature is the visible party rather than the lab that trained the model.
- contradiction The researchers closest to the technology both warn it could end the human race and sit on the 15% growth side of the bet, while the economists in the same post forecast slow diffusion. Industry forecasts cannot anchor the risk case or the upside case.
- decision With opinion moving and no legislation described, the live choice is about the assumptions inside a two-year deployment plan.
The opposition is five separate interests. For anyone forecasting how long it holds, the composition matters: a liability safe harbour that satisfies investors does nothing for a China hawk who wants the race won faster, and a libertarian who objects to regulating technological progress on principle is unmoved by the administration's growth argument [3]. The roundup called the group "a strange alliance indeed" [4].
The evidence for the turn in opinion arrives secondhand. The post credits Nate Silver with rounding up the polling, and does not name the pollsters or give the figures [6]. On that basis its author wrote: "I don't think I've ever seen such bipartisan national unity on any issue in my adult lifetime" [7].
The second half of the same post sets up a harder problem for planners. The public bet it describes has one side forecasting 15% growth [10], against the 5.3% that AI experts gave as their most optimistic scenario [11]. The 15% forecast is about 2.8 times that optimistic case [15]. The post notes China has hit 15% in only one year since its rapid growth began, 1984, and it was a very poor country then [12]. The fast-growth side is mostly AI researchers and the slow-growth side mostly economists [16].
Alex Imas and Ben Moll, cited in the post, name four brakes on AI's short-term contribution. They are slow diffusion through the economy, the J-curve dip while firms spend on adoption instead of their existing businesses, the difficulty of automating the physical world with robots, and political barriers to adoption [13]. That last item connects the two halves of the roundup. If majorities want AI slowed, the first place a company meets that is inside its own adoption decisions, in procurement reviews and customer questions, well before any statute exists.
The post itself puts the president on the other side of this one, and disputes how long that lasts. It writes that Trump is standing firm against the tide of public opinion, as he has on the Iran War and tariffs, and that "it's not clear how long he'll be able to hold out" [14].
For a decision taken this quarter, the distance between a bipartisan majority and an enforceable rule is a legislative process, and this account describes opinion. The cheaper adjustment is to the deployment plan itself. Write the safety claims so the engineers who shipped the feature can defend them to a customer or a committee.
What to watch
- Whether the polls behind Nate Silver's summary are published with question wording and fieldwork dates. That would let a planner size the majority.
- Whether Democrats adopt AI safety as a campaign plank after Obama's advice, the first observable step from opinion toward legislation.
- Whether the progressive split between the "AI is dangerous" and "AI is fake" camps resolves, since that decides whether restrictions arrive with sponsors on both sides.