There has been a lot of talk about “AI doom” or “AI crisis”. My position is the same now as it was years ago: despite exponential progress in AI, society remains mostly unchanged: No mass-unemployment, no collapse of social order. Over the longer-term, productivity goes up, but society itself does not change structurally aside from AI playing a large role in it, much like smartphones or personal computers.
As of September 2026, AI has not produced mass unemployment in the United States. Predictions of crisis or doom often amount to goalpost moving, by changing the timeframe or conditions when the predicted crisis fails to materialize. Other times, the predictions are vague or worded in such a way as to preclude the possibility of ever being wrong.
As for specific predictions, I asked Astra about AI 2027’s job-loss forecasts, but it didn’t give a definitive answer:
AI 2027’s forecasts are so vague as to be useless, so as to avoid being wrong or having to commit to a side. In any dynamic economy, jobs are constantly being created and destroyed. What matters is the actual net job creation/loss as indicated by “official” BLS stats (e.g., the unemployment rate or long-term unemployment rate, which are conspicuously absent from 2027’s forecasts).
I then asked for a specific number…nothing:
Yes, the BLS stats are far from perfect and inaccurate for certain categories, but they’re as close as we’re going to get to anything objective for tracking this sort of stuff. The narrative for the last 3 years is a net loss of jobs due to AI, not merely jobs being shuffled. So far, this has not happened.
This may come off as dismissive to say that the opposite side is wrong, but they have yet to provide anything resembling specifics. When you press them on the question of, “What must transpire for AI to destroy humanity?,” they deflect by invoking certain thought experiments; e.g., likening it to playing chess against a grandmaster, that being AI. Take, for example, this tweet:
It is true. I cannot point to a single instance of AI killing everyone. I should wait until I can point to such an example, some people say. https://t.co/86IKxgYeXo
— Zvi Mowshowitz (@TheZvi) September 17, 2026
It sounds clever and is convincing to a lot of people. Yes, if our choices are the destruction of humanity or pausing AI, then obviously we should pause AI. But such a framing is not an argument. It presupposes or assumes what needs to be proved: that the continued development of AI means the end of the world. No evidence is ever supplied.
But the thing is, the pause or slowdown is already underway. His side is winning. Everyone, save for Trump and some random people online, is saying to pause or slowdown AI. Although they position themselves as outsiders against the “AI hegemony,” their position is the mainstream. It’s very hard, if not impossible, to find a respectable journalist or industry expert pushing for AI acceleration. Hell, even the AI CEOs themselves are begging to be regulated and paused.
But overall, it’s more intellectually honest to say, “Yeah, we have no idea what will happen,” than to make vague forecasts that cannot be proven wrong or present false dichotomies that make the sought answer seem inevitable. If you think AI will kill the world or cause mass unemployment, your job is to work through the logical flow of how that happens, not leave it up to the skeptic to do your intellectual heavy lifting for you.