There is a particular kind of post making the rounds right now. Someone feeds an old and ominous text into a chatbot, the Book of Revelation, say, or Nostradamus's quatrains, asks it to find the hidden pattern, and reports back that the machine has "decoded" a date for some coming catastrophe. The date is almost always 2027. The framing is always that this is evidence, not entertainment.
It is worth saying plainly. None of this is real, and the AI is not doing what the post claims.
Why it feels convincing
The genre is old. People have been finding doom in Nostradamus and numerical codes in scripture for centuries. What is new is the prop. Slotting an AI into the process makes the same activity feel rigorous, because pattern-matching done by a machine reads as more objective than pattern-matching done by a person, even though it is the identical exercise. The chatbot lends an air of computation to what is really just interpretation.
Here is the mechanism people miss. A large language model generates text that plausibly continues your prompt. Ask it to find a hidden apocalypse date in an ancient text and it will find one, because that is what you asked for. Ask the same model to prove the same text points to 2031 instead, and it will happily oblige. It is not consulting hidden knowledge. It is producing the shape of the answer you requested, dressed in confident language.
Real forecasts look different
None of this means every claim about 2027 is nonsense. There are serious, detailed forecasts about where AI capability might be heading, and some of them do point at the next few years. The difference is that those forecasts are falsifiable. They make specific predictions you can check against reality, and by most accounts the well-known ones are running roughly two-thirds on pace. A decoded prophecy makes no checkable claim at all. It cannot be wrong, which is exactly why it should not persuade you.
The anxiety underneath is understandable. When a lab founder like Dario Amodei is publicly urging the industry to slow down, it is easy to reach for older languages of doom to make sense of the moment. Fair enough. Just do not mistake a chatbot echoing your fears back at you for a machine that has seen the future. If you want to worry about AI, there are better and worse things to worry about, and the test is always the same. Can the claim be checked?
Commentarii · 0