|
randombio.com | Science Dies in Unblogginess | Believe All Science | I Am the Science Monday, March 23, 2026 | science & philosophy Is ‘emergent’ artificial intelligence possible even in theory?Emergence is not a real phenomenon, but a placeholder for something not understood |
umans sometimes worry that intelligence could
emerge spontaneously from a collection of unintelligent parts
like cell phones. It’s a common fear that chatbots, which use
neural network algorithms, could somehow acquire consciousness
and take over the world. Some claims, like those in
popular books,[1]
have become quite hysterical. These authors take it for granted that
a super-AI that wipes out the humans is inevitable.
That would be a darn shame. But contrary to popular opinion, it won’t happen by emergence. An emergent property is one that occurs spontaneously and unpredictably and is only apparent in a large collection of parts. It does not mean that the property doesn’t exist in the individual part, only that it can’t be measured because the property is evident only when the parts interact.
Emergence is a metaphysical question like the question of whether a bent wire is the same wire as the original straight one. Like many questions in metaphysics, resolving it would have no practical consequences whatsoever. But it’s helpful to consider it to dispel irrational fears.
Some properties are generally considered emergent. Pressure, for instance, is most meaningful when there are many individual particles in a confined area, though a single molecule does exert a pressure. Unlike acidity or temperature, pressure is hard to measure for an individual particle. Wind speed might be another one. Emergence is also still used in quantum gravity, where the idea is that spacetime somehow ‘emerges’ from something more fundamental. To be emergent, that ‘something’ would have to exist but have no duration and no size. Imagining how that could work is a major challenge.
But for the most part, emergence is too vague to be a useful scientific concept. Emergence is a placeholder, like dark energy. Once the phenomenon is understood, ‘emergence’ is replaced by something else: a causal explanation.
Brain function, for example, is not an ‘emergent’ property. If it were, we would have to conclude that anything made of parts, such as a radio or a piano, was exhibiting emergent behavior.
A device like a radio or a brain or a piano is built from parts and has the properties imparted by its inventor or by evolution. Even though there is no ‘pianoness’ in the thousands of wooden bits and pieces in a piano, the music made by a piano is not an emergent property but an engineered one. It is no more emergent than the percussive noise caused by (heaven forbid) hitting a piano with a baseball bat. Clearly, ‘piano-banginess’ is not an emergent property, and indeed not a bona fide property at all.
Hitting a rock with another rock is a good example. You throw a rock. It hits a second rock and moves it. That can’t happen with a single rock. Is that emergent behavior? Maybe the rock fell spontaneously from the sky and hit the second rock. Do we have the new emergent property of “rock hittiness”? Clearly emergence is not needed here, because we fully understand how one physical object causes another to move. It might occur spontaneously, but it is far from unexpected.
Chemical properties are also not emergent. Two chemicals can react with each other spontaneously, but this is not really an emergent property. It is a predictable and well understood result of the electronic and chemical properties of the molecules. A chemist can easily predict what two chemicals, such as formaldehyde and ammonia, will do when they are mixed together.
Another non-example is a flock of birds. While it is true that a ‘flock’ is an aggregate property, flockiness is not unexpected, since we already know that any large group of birds constitutes a flock. The only difference is the conceptual one. What is needed for emergence is not dependence on parts, but a completely new and unexpected property that arises in the aggregate.
So, getting back to our original example, can cell phones or chatbots exhibit emergent intelligence? People who study the brain, aware of the elaborate architecture and the thousands of specialized functional parts it depends on, would say the idea is preposterous. Such a thing would require deliberate engineering based on an understanding of how intelligence works, which the humans don’t have. A chatbot could be engineered to be intelligent if the humans knew how to do it. But the idea that it could happen accidentally is only a way of generating scary headlines.
Consciousness is also not an emergent property. To the best of our knowledge, consciousness consists of two features: perception / awareness, and information internality / subjectivity. Perception is no more ‘emergent’ than a radio is ‘emergent’ from a collection of capacitors, resistors, and transistors. To be emergent, the property would have to ‘emerge’ spontaneously and unexpectedly. This is clearly not something that happens in a pile of electronic parts unless they somehow started receiving Classic Rock 97.1 without needing to be assembled and powered up.
Rickabaugh and Moreland [2] distinguish emergentism from early spiritualist ideas like substance dualism that conceive of consciousness as a simple substance, which means it has no individual parts. Consciousness, in this view, is like a magical liquid that when poured into your head changes you from a zombie into a conscious being. In this idea, consciousness was even thought to be able to exist without a physical body.
They cite philosopher Robert Koons who argued that emergence of consciousness is unlikely:
Emergence would require a very improbable and ad hoc preestablished harmony among the powers of many mutually unifying parts. [3]
The idea of substance dualism as an explanation for consciousness is no longer widely held, but we can agree with the dualists that emergence is not a real thing, but only a placeholder for as-yet undetermined causal effects.
An ocean, for example, is not emergent even though its properties such as waves and wetness depend on how water molecules interact with each other. Somehow the question of whether water is wet or not has turned into a big debate. But we know that the unbonded electron pair on water determines its shape and its hydrogen bonding, which induces properties like high boiling point when many water molecules flock together.
We might say ‘wetness’ emerges from a swarm of water molecules, but we would never say the ocean ‘emerged’ from them because the idea would be ridiculous. Yet we still talk about oceans, pianos, and flocks of birds. The answer is plain: emergence is all in your head.
[1] Yudkowsky E, Soares N (2025). If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All, Little, Brown
[2] Rickabaugh B, Moreland JP (2024). The Substance of Consciousness. Wiley-Blackwell, p. 39.
[3] Koons RC (2018). in The Blackwell companion to substance dualism, p377–393.
mar 23 2026, 8:04 am
The Age of Hysteria
“If anybody builds it, everybody dies.” Does anybody
really believe that? Of course not
The Turing Test is worse than useless
Chatbots outperform humans on the Turing test. Tweaking the test
until it gets the right result won’t help
Can artificial intelligence ever be conscious?
Brain science says yes, religion says no. Just don’t
confuse a chatbot with something intelligent