Skip to content

On Aphantasia and LLMs or: Which Human Are We Talking About?

A few years ago, I discovered I was completely and profoundly aphantasic. Like almost all aphantasics, I discovered it by accident, as an adult, stumbling on an article that described the condition. I expect that around 5% of the readers of this piece will discover, today, for the first time in their lives, that they are aphantasic like me. And from today, their lives will change, as mine did.

Aphantasia was first described in 1880, only to vanish from human knowledge until 2010, when Adam Zeman, an English neurologist now at Exeter, rediscovered it by chance in a patient who had undergone coronary angioplasty. A few days after the operation, patient MX went to bed one evening like any other and woke up completely unable to visualise anything in his mind. His visual imagination, once vivid and central to him, had disappeared entirely. MX could see normally — but only what was in front of his eyes.

Zeman wrote a frankly semi-obscure case report on the matter which was picked up by chance by the science journalist Carl Zimmer for Discover Magazine, an American popular science publication. Some of the readers of Zimmer’s piece recognised themselves in the description of patient MX, but with one fundamental difference: they had never had the capacity to form visual images in their thoughts, and had been born that way. Their inability to visualise was congenital, and the very idea that some people could do it left them perplexed. Contacted by around twenty such readers, Zeman realised the phenomenon must not be so rare, and in 2015 he published a second paper describing congenital aphantasia in scientific terms for the first time.

There are various tests to determine whether someone is aphantasic, and to try to give it a measure (as always happens in biology, this too is a continuous spectrum, not a binary). The following is my favourite:

Imagine a room with a table inside it. On the table there is a piece of fruit, and next to the table there is a person wearing a hat. Now I ask you: what shape is the table? What fruit is on it? Was the person a man or a woman? Young or old? What colour was the hat?

Most people — an estimated 90% — can answer all of these questions, because their heads have literally hallucinated the answers. Most people, that is, can tell you whether the table was square or round, whether the fruit was a banana or a watermelon, and so on. In their minds, an image has appeared — a mental hallucination — which has automatically filled in all those gaps, even adding independent details as well: wallpaper, curtains, glasses, carpets. Some people instinctively build a story around that image. A friend could tell me everything not just about that man and his hat, but all his family too.

A total aphantasic like me cannot answer any of those questions, because that information simply does not exist. Once you grasp this limitation, not only do the questions stop making sense: the very idea of “imagining” that scene stops making sense. Usually, an epiphany follows, and you realise that the primary-school teacher who asked you to “imagine” a house by drawing it was not speaking in metaphor.

Important to point out that, since aphantasia is a spectrum, some people will have a weak or incomplete image. They might be able to say whether the person was a man or a woman, for instance, but not describe them.

In general — though not always — an aphantasic also loses other common capacities: the ability to recreate a voice or a piece of music internally (anauralia); the ability to have an inner voice (anendophasia); or, like me, the capacity to recreate or recall any sensory information at all (multisensory aphantasia): no images, no voices, no smells, nothing.

If I close my eyes, I enter what you probably call meditation. Everything disappears.

Although it is a condition and not an illness, aphantasia statistically leads to some consequences in real life. Aphantasics tend to be less inclined toward the arts and more inclined toward disciplines that demand abstract thinking (the proportion of aphantasics is slightly higher among STEM people). Aphantasics are often immune to PTSD, being unable to re-see or re-live past images. An aphantasic is also unable to re-see the face or re-hear the voice of loved ones, and so tends to experience bereavement in an apparently simpler way, and grief more transiently — and yet, at the same time, this remains one of the greatest regrets of those who discover they are aphantasic, along with the fact of being unable to imagine anything when reading a novel (I only read non-fiction). These conditions are also often associated with a total inability to recall autobiographical memories (SDAM: Severely Deficient Autobiographical Memory) — the inability to re-live one’s own past.

The way I live and exist is essentially different from that of 90% of the population. Your conscious experience is different from ours in a way that is almost difficult to conceive. I have no visual memory, only conceptual memory. I can tell you the colour of the shirt my son was wearing this morning, just as I can tell you the colour of the sky, or the chemical formula of water, or the structure of glutamate. Not because I visualise those things, but because I simply know them — just as I know that my name is Giorgio and where I live.

Now, given my profession and my personal interests, my experience has been an absolutely unique and privileged vantage point from which to approach the entire literature on consciousness. My concept of consciousness is profoundly different from yours, as is my experience.

When I hear people mock an LLM because “it hallucinates”, I find myself as the odd one out because we aphantasics have never had hallucinations, and we are, in the end, the only ones who can truly say we think without hallucinating. Everyone else does it constantly — they just call it remembering, imagining, dreaming.

When people say that an AI cannot function like a human, my instinct is: which human? Which human are we talking about? Me? You? Or are we taking for granted that all humans share the same cognitive experience? It took thousands of years to give a name to a phenomenon like aphantasia. Who knows how many other aspects of the human mind each of us takes for granted as universal — when they might be rare, or even unique.

The epiphany of aphantasia can hit different people in different ways. Most people who are able of mental imagery have a hard time trying to imagine (is that even the right word now?) what life would be like without the ability to create visual representations or recall biographical memory or close one’s eyes only to deepen into sheer emptiness. How can these people work? The answer is: we just do. There is only one way to judge and measure consciousness and intelligence and that is functionalism. I talk about functionalism in my most recent preprint on intelligence and AI. Worth a read, here.

Leave a Reply

Your email address will not be published. Required fields are marked *