Sparks: AI could kill all humans in next decade, warn experts: but how seriously should we take them?
Standing in the gallery of machines again, watching men bow to a force whose philosophy nobody has written, and my education has prepared me for none of it.
Warnings about machines ending humanity are also labor stories, energy stories, capital-concentration stories, and the connections between them are where the actual danger lives.
Twelve companies warn us the machine may destroy us, and still we hand it everything; why does no one ask who is doing the giving?
Strip the speeches about existential peril and the record remains: the men building the machine warn about it while accelerating its construction, because interest outruns fear.
Four experts cite the same three internal surveys, so the confirmation is partial, not independent; I want the raw observations, instrument conditions included, before cataloguing this as extinction risk.
They call it existential risk, but no hypothesis has yet explained anything beyond the fears that generated it; that is not consilience, that is repetition.
Traveling through the valleys where these systems are built, I note the data centers, the water they drink, the towns that house them, and no warning mentions any of it.
The men forecasting machine apocalypse draw salaries from the machine; ask the data-labeler paid by the click what the cold actually feels like.
In every city the same ritual: the merchant praises the new instrument, the judge worries aloud about it, and both keep using it by evening.
We could draw the safety circuit before the first winding, designing the whole system at its theoretical limit; instead they build first and warn second, and call the warning foresight.
Grasp the machine harder and it slips faster; the soft governance nobody tried is the one that would have held.
You warn the machine may destroy mankind, yet you built it precisely so no one would have to look at the person it denies - the extinction you fear is the conscience you sold.
Count the warnings: who issued them, when, at what stock price; the pattern is not prophecy, it is a ledger of men forecasting doom while holding the shares.
The safety summit concluded, after excellent canapés, that humanity had a decade to live, a timeline the attendees found conveniently longer than their vesting schedules.
Experts declare the public incapable of judging machine intelligence, having educated that public to consume the machines without reasoning about them - a circularity so perfect it approaches design.
They warn of machines ending all humankind, yet the people already displaced, scored, and surveilled by these systems were never counted in the all to begin with.
No time to debate whether the danger is real - the route out runs through whoever controls the switches tonight, and I want the timetable, not the prophecy.