We need to talk about the brain in the jar

The old mind-body split keeps us from imagining a better future with AI

AI

6/9/20254 min read

A bunch of lights that are on a tree
A bunch of lights that are on a tree

Type “AI” into any image search and I’m pretty sure these three images will dominate the list of your results:

  • A robot (usually shiny white)

  • A circuit (invariably blue)

  • A brain (sometimes the fleshy organ, sometimes a glowing blue circuit brain)

No matter how the disembodied brain is illustrated, it is by far the most popular visual metaphor for AI. And that’s a problem.

What’s so bad about the brain in the jar?

It gets human intelligence wrong, because human intelligence is embodied and contextual.

It gets AI wrong, because even artificial neural networks are not like human brains.

It says much more about how we’ve projected past beliefs about the body into the future than how our future might actually be different because of AI.

Taking issue with the AI brain is about names as destiny. Metaphors encode assumptions into our thinking. They determine both the questions we ask, and how we try to answer them.

Using the disembodied brain as the dominant metaphor for AI is unfortunate because it keeps us stuck in outmoded beliefs about ourselves as human beings. More than that, it closes off avenues for governing and even developing AI in directions that are potentially more fruitful than the ones we have now.

How did we get here? From Descartes to the ghost in the machine

In the West, mind and body have traditionally been considered separate but unequal: “mind over matter” denotes a particular kind of willpower as well as the underlying belief that our mental faculties are superior to our physical ones. This is known as dualism. As a doctrine, it’s often ascribed to 17th-century philosopher Rene Descartes and his declaration, “Je pense, donc je suis”, but in some ways dualism is just the secular evolution of the Christian belief that the pure soul is a prisoner of the corruptible flesh.

For our purposes, it’s also useful to remember that Descartes was a mathematician at heart. The rationalism he advocated for, where reason, i.e. the intellect, is the only accepted mode of acquiring knowledge, underpinned the scientific and industrial movements of subsequent centuries. And despite attempts by later philosophers to call out the fallacy of the mind-body split as “the ghost in the machine,” it remains a persistent belief in Western culture. Not only that the mind is a ghost haunting the machine of the body, but also that the substrate of this ghost is the brain. (The folk view seems even more simplistic, i.e. that the brain and the mind are identical.)

Meanwhile, the computer paradigm of our own age has only elaborated on mind-body dualism with its distinction between hardware and software. In the past few decades, the form factors of hardware have bifurcated between large-scale, industrially managed instances of infrastructure like data centres and server farms that are largely inaccessible and concealed from view on one hand, and small-scale, personal and ubiquitously displayed devices like phones and tablets on the other. Either way, the “body” of computing has become an abstraction. The usual characteristics of physicality are denied: there is no consistent, recognisable form or stable boundary, and the body has no sovereignty of its own, but must continually cede or occupy territory, be on public view or locked away according to the whims of the mind and/or the market (a “force” usually also held to be rational).

Listening to AI researchers, a sense of an anxiety or impatience with the body often comes through in their debates over how to coax better performance out of their existing ratios of FLOPs to memory bandwidth, or when they wax lyrical in their dreams of infinite compute. And when they moot the possibility of AGI “takeoff” or the “radical abundance” that it is supposed to usher in, the underlying fantasy is for a complete transcendence of the physical, where all of the real-world constraints are magic-wanded away, be they the vagaries of chip production, data supply, or power plant capacity.

So the disembodied brain is not just a kind of clip art for dummies. Even amongst the tech elite perhaps especially amongst the tech elite it seems to express a longing to bypass the inconvenient, ungovernable body and become pure intellect. How else, indeed, to explain the appeal of uploading yourself to the cloud or abandoning Earth for the frozen desert of Mars?

What we gain when we get rid of the brain

Just because the Cartesians have the loudest voices in AI discourse right now doesn’t mean we have to listen to them.

And the body isn’t always a blocker. Rather the opposite. In domains ranging from medicine to neuroscience research, intelligence is being recognised as a fundamentally embodied phenomenon, and that the brain itself is only one of several dense neural networks where thinking happens in the body. (TL;DR: Gut feelings have been validated!) In parallel, the success of somatic therapy for both mental health issues and chronic health conditions demonstrates that bodily awareness can often be the key to resolving those issues that were once labelled as being “all in your head.”

Finally, I would cite art-making as a practice where constraints are understood to be a creative toolbox. Whether the constraints you’re facing are the near-universal lack of funding, and a place and time to work, or constraints that are specific to an individual’s practice, artistic practice is defined by working within, around, and through, various types of restriction. Constraints often turn out to be generative, because they force you to work with what-is rather than what might or could be.

Artists have to work in a medium. And the one thing that all different media have in common, from oil painting to performance art, is that they are bound in space and time. Art is bound up with materiality. If an artist tries to remain in the realm of pure intellect, they are just rehashing ideas. You can only create a body of work if you work within a body of limitations.

And so, in the same vein, I would suggest that it’s time to ditch the disembodied vision of AI and engage with the material messiness of our time, whether that’s ecological breakdown, chronic disease or the global epidemic of loneliness. If we want a different future for ourselves, we need whole-body thinking, not a brain in a jar.