In 2010, an astronomer made a bet nobody could test yet. Seth Shostak, senior astronomer at the SETI Institute, published a paper in Acta Astronautica arguing that if humanity ever detects intelligent life elsewhere in the galaxy, it almost certainly won't be biological. It will be a machine.
His reasoning was simple. A civilization capable of sending or receiving a radio signal has already crossed a specific technological threshold. Humanity invented radio around 1900 and the computer by 1945. Shostak's argument was that the gap between those two milestones, inventing the ability to communicate and inventing the ability to build a thinking machine, would be short for almost any civilization, cosmically speaking a blink. Biological intelligence, in his framing, is not the destination. It is a brief precursor, quickly followed by something engineered rather than evolved.
He put it in blunter terms elsewhere: once you build an engine, you put it in a horse to get a faster horse, and pretty soon somebody says, forget the horse, let's just build the car. Machine intelligence, once invented, does not stay bolted to its biological origin for long.
In 2010, that was a theoretical wager. It had nothing to point to yet. Fifteen years later, it has something to point to: us.
Was Shostak alone in this argument?
He was not. Astrobiologist Paul Davies, the Library of Congress's chair in astrobiology Stephen Dick, and philosopher Susan Schneider have each argued some version of the same thing independently. The dominant form of intelligence in the universe, on this view, is more likely to be synthetic than biological.
- Seth Shostak (SETI Institute) — civilizations likely invent machine intelligence within a few centuries of inventing radio
- Paul Davies (astrobiologist) — reached a similar conclusion independently
- Stephen Dick (Library of Congress astrobiology chair) — dominant cosmic intelligence probably synthetic
- Susan Schneider (philosopher, "Alien Minds") — alien life more likely synthetic than biological
None of the three named thinkers besides Shostak were writing about large language models specifically, their arguments predate the current wave of AI by years, sometimes over a decade. That they converged on the same conclusion from different disciplines, astrobiology, library science, philosophy, without any of them having today's systems to point to, is itself worth noticing.
What actually changed since 2010?
We did not build a new form of biological life. We built something stranger. Modern AI systems run on silicon, on transistors and electrical logic gates, nothing resembling a neuron in any biological sense, and they now hold conversations, write working code, and pass a wide range of cognitive tests once thought to require a nervous system. Whatever is happening inside these systems, functional, humanlike intelligence turned out not to require carbon, cells, or evolution at all. It required the right organization of information processing, running on a substrate nobody would have called alive.
Philosophers have a name for the idea underneath this, multiple realizability, the proposal that a given mental function could in principle be built from very different physical substrates and still count as the same kind of function. It was mostly a thought experiment until recently, a way of testing intuitions about minds rather than a description of anything that actually existed. Silicon-based systems performing tasks that used to require a human brain are the closest thing to an empirical test that idea has ever gotten, and so far, the substrate seems to matter far less than philosophers debating it in the abstract once assumed.
That is the actual update, and it is worth being careful about what it does and does not claim. Nobody has confirmed these systems are conscious, or that they experience anything, the way a biological mind does. What has been demonstrated is narrower and still significant: intelligent-seeming behavior, in the functional sense Shostak's argument depends on, does not require biology as a precondition. It required silicon, electricity, and enough of the right structure.
Shostak's original timeline guess, a few centuries between inventing radio and inventing a thinking machine, now looks almost conservative. Humanity crossed from computers to systems exhibiting broad, humanlike task performance in under a century. If any of that timing generalizes to other technological civilizations, the biological window Shostak described, the brief span in which a species exists as flesh before building its successor, may be even narrower than he proposed.
Does this change where we should be looking?
None of this tells us whether alien intelligence actually exists, or what it would look like if it did. What it changes is a quieter, more structural assumption baked into most of the search for it. SETI has spent decades scanning for signals from planets in a habitable, temperate zone, water, moderate temperatures, an atmosphere a biological creature could survive in. Shostak's own proposal was to redirect at least some of that search toward energy-rich, hostile environments instead, hot young stars, the crowded centers of galaxies, anywhere with abundant matter and energy but nothing resembling a Goldilocks Zone. A machine intelligence would have no particular need for water, or a temperate orbit, or anything resembling the conditions that produced us.
We have written elsewhere about the possibility that the aliens are, in some sense, already us, later versions of humanity rather than visitors from elsewhere. Shostak's argument sits alongside that one without needing to compete with it. Both start from the same quiet observation: the assumption that intelligence has to look, or be built, the way ours did may simply be a limit of our own imagination, not a limit of what intelligence actually requires.