What Do You Dream When the World Goes Silent?

Close your eyes.

Picture a forest.

Go ahead — take a second.

What arrives first: a smell, a quality of light, the sound of something moving through undergrowth, the specific texture of bark under your palm? Or does a stock photograph materialise? A cinematic establishing shot, colour-graded in that particular teal-and-orange palette that streaming platforms have trained us to associate with nature? The question is not rhetorical. It is, arguably, the most consequential diagnostic of our moment. Because the images that arrive when you close your eyes — the raw material of your imagination, your dreams, your interior world — are now, for the first time in human history, substantially sourced from outside you. That should unsettle everyone in this room.

I. The Flood and What It Drowns

The average human being in 2024 encounters somewhere between 6,000 and 10,000 visual stimuli per day — advertisements, thumbnails, reels, AI-generated images, algorithmically curated short-form video — a figure representing an increase of roughly 5,000% over the estimated exposure of a person living in the 1970s.^[1] This is a measurable ecological change in the informational environment in which human cognition develops, operates, and — during sleep — attempts to restore itself.

The anthropologist David Howes, whose work on sensory culture repositioned the senses as culturally constructed rather than biologically fixed, argued that every culture trains its members to perceive in particular ways — to attend to certain sensory channels and suppress others.^[2] What Howes could not have fully anticipated was the speed at which a single technological shift could reconfigure that training at civilisational scale. We are living inside that reconfiguration. The human visual cortex — which processes images roughly 60,000 times faster than text, as documented in MIT neuroscientist Mary Potter’s rapid serial visual presentation studies^[3] — is now receiving more pre-fabricated imagery per waking hour than it received per waking year two generations ago. The cortex processes it all. It stores what it is given. It dreams from what it has stored.

• Avg. daily screen time globally (2025): 6 hrs 37 min / person

• Short-form video consumption: up 85% since 2020

• AI-generated images produced daily (est. 2024): 34 million+

• Reading without image assistance (books, longform): down 57% among 18-34s since 2010

The psychologist Daniel Levitin, in The Organized Mind (2014), described information overload as producing what he called “neural fatigue” — a condition in which the prefrontal cortex, responsible for prioritising and filtering experience, exhausts itself managing an environment of perpetual novelty.^[4] What Levitin framed as a productivity problem is, from an anthropological vantage, something more foundational: a crisis of imaginative sovereignty. When the brain is perpetually managing incoming images, it has fewer resources for generating them. The default mode network — the neural architecture responsible for mind-wandering, self-referential thought, and crucially, the consolidation of memory during rest — becomes increasingly colonised by what has been consumed rather than what has been imagined.

II. The AI Image and the Death of the Inner Picture

Here is where the argument sharpens into something that should genuinely alarm us. The images now flooding the visual ecosystem are, in rapidly increasing proportion, generated by artificial intelligence. As of 2024, Adobe, Midjourney, DALL-E, and Stable Diffusion collectively produce an estimated 34 million images per day — a figure exceeding the total output of professional human photographers globally.^[5] These images share a family resemblance that trained eyes can identify: a particular softness of light, an almost-but-not-quite-right quality of hands and faces, an aesthetic consistency derived from training data rather than from any act of human seeing. They are, in the most literal sense, images that no one has ever seen with their eyes. They are statistical aggregations of human visual output, recomposed by a system that has processed more images than any human being could encounter across a thousand lifetimes.

The AI image is the visual equivalent of a sentence generated by averaging every sentence ever written. Legible. Often beautiful. And never, not once, surprised by the world.

The philosopher and art critic John Berger, in Ways of Seeing (1972), made the foundational argument that seeing is an act — that the way we look at things is shaped by what we know, what we believe, what we have experienced.^[6] To see, for Berger, was always an interpretive and therefore irreducibly personal act. The AI image reverses this relationship entirely. It is produced by a system that has seen everything and experienced nothing. It flattens the space between image and viewer, offering a product so visually saturated, so immediately pleasing to cortical pattern-recognition systems, that it bypasses the slower, more effortful process of meaning-making that genuine looking requires. And yet we scroll through these images by the thousand. Our visual cortices accept them. Our hippocampi file them. Our dreams, eventually, are furnished with them.

III. What Reading Does That Watching Cannot

There is a cognitive argument for reading without image assistance that the current discourse has catastrophically failed to make clearly enough, so let us make it here with some force. When you read the sentence “the river ran the colour of old copper in the last hour of light”, your brain does something that no screen, regardless of resolution, can do for it: it constructs an image. That image is yours. It is assembled from your specific archive of rivers seen, of copper objects handled, of particular late afternoons lodged in memory. It is irreducibly personal. It is, in the most literal neurological sense, an act of creation.

Maryanne Wolf, the cognitive neuroscientist at UCLA and author of Reader, Come Home (2018), has documented what she calls the “reading circuit” — a network of neural regions that, when engaged with text without visual accompaniment, activates language-processing areas alongside the motor cortex, the olfactory cortex, and the regions associated with emotional simulation and theory of mind.^[7] Reading, Wolf argues, is the most cognitively demanding act of imagination available to the human brain in ordinary daily life. It is the gymnasium of the inner life. And it is in measurable, documented decline.

Between 2010 and 2024, the proportion of Americans who read for pleasure for more than thirty minutes per day fell from 28% to 14%.^[8] Among 18–34 year olds, the decline is steeper. The replacement activity is, almost universally, short-form video consumption — a form of media that, by design, pre-fabricates every visual element and delivers it at a pace calibrated to prevent the attentional wandering that genuine comprehension requires. The brain that watches is receiving. The brain that reads is building. These are physiologically distinct operations with measurably different consequences for the architecture of cognition, and treating them as equivalent activities dressed in different clothes is one of the more expensive intellectual evasions of our time.

IV. What It Does to Children

The developmental stakes of this shift are, when examined with any care, alarming. Jean Piaget’s theory of cognitive development — still the most comprehensive framework available for understanding how children construct their understanding of the world — describes the period between ages 2 and 7 as the preoperational stage, characterised above all by the emergence of symbolic thought: the capacity to let one thing stand for another, to form mental representations, to imagine.^[9] This is the developmental window during which the inner picture, the personal image-archive, is being built. It is also, in 2026, the window during which the average child in a high-income country is already accumulating over three hours of daily screen exposure.

The developmental psychologist Sandra Calvert at Georgetown University has documented that children under five who consume high volumes of fast-paced visual media show measurably reduced capacity for what she terms “mental imagery generation” — the ability to construct vivid, self-authored internal representations in the absence of external visual stimulus.^[10] In plain language: children shown too many images too early in development struggle to picture things on their own. The imaginative muscle, like any other, atrophies without exercise.

We are raising the first generation in human history whose imaginative vocabulary was authored, in substantial part, by someone else — specifically, by a machine trained on the aesthetic preferences of advertisers.

The consequences extend beyond imagination into empathy and social cognition. Sherry Turkle, MIT sociologist and author of Reclaiming Conversation (2015), has documented with longitudinal rigour the correlation between reduced face-to-face interaction — driven by device dependence beginning in early childhood — and measurable declines in empathic response and perspective-taking capacity.^[11] Empathy, Turkle argues, is built through the slow, ambiguous, often uncomfortable experience of sitting with another person’s face and attempting to read it. AI-generated facial imagery provides a version of faces from which all genuine ambiguity has been engineered out. The child who grows up reading AI-smoothed faces is being trained on a dataset that has removed precisely the information — the micro-expressions, the asymmetries, the unguarded moments — that empathy requires.

V. What It Does to Adults

Adults, of course, are implicated in equal measure. The assumption that a formed adult brain is insulated against environmental reconfiguration is one of the more comforting and less accurate beliefs of popular neuroscience. Neuroplasticity — the brain’s capacity to reorganise its structure in response to experience — does not cease at 25. It continues, with diminishing elasticity, throughout life. And the experience the adult brain is currently receiving, at unprecedented volume and velocity, is pre-fabricated visual content authored by algorithmic systems optimised for engagement rather than enrichment.

The sociologist Georg Simmel, writing at the turn of the twentieth century on the sensory experience of metropolitan life, described the condition of the urban dweller as one of Reizüberflutung — stimulus flooding — and argued that the metropolitan personality adapts through what he called the blasé attitude: a psychic self-protection mechanism involving the flattening of affective response to the world.^[12] Simmel was describing the relatively modest visual overstimulation of early twentieth-century Berlin. His framework, applied to the contemporary scroll, describes something that makes Berlin look like a meditation retreat.

The blasé attitude of the chronically online adult is so thoroughly normalised — so completely memed into a personality trait — that we have lost sight of what it represents clinically: a measurable reduction in the capacity for wonder, for sustained aesthetic attention, for the kind of slow looking that produces genuine encounter with the world. The art historian T.J. Clark, in his late essays on slowness and looking, argued that the encounter with a great painting requires a minimum of twenty minutes of sustained attention before it begins to yield its meanings.^[13] Twenty uninterrupted minutes. The average duration of engagement with an image on social media is 1.7 seconds. We are processing. Calling it looking is a category error.

VI. The Dreaming Mind and Its Raw Material

This is where the argument arrives at its most visceral and least discussed territory: the dream. Sleep researchers have, over the past three decades, developed a sophisticated understanding of the relationship between waking experience and the content of REM dreaming. The neuroscientist Matthew Walker, whose Why We Sleep (2017) brought sleep science into general consciousness, summarises what is now a well-established position: REM sleep functions as a form of overnight emotional and experiential processing, during which the brain replays, recombines, and integrates the material of waking life into the architecture of long-term memory and self-narrative.^[14] What you dream from, in other words, is what you have lived with.

The psychologist Ernest Hartmann, whose theoretical work on dream imagery proposed that dreams function as a form of contextualisation — placing the emotional concerns of the dreamer within a narrative frame that reduces their psychological charge — argued that the imagery of dreams is drawn from the totality of the dreamer’s experiential archive, weighted toward material carrying unresolved emotional significance.^[15] The dream selects from what is available. It assembles its arguments from the images the waking mind has accumulated. The question that follows from this, with a logic that is almost mechanical: what happens to the dream when the majority of its raw material is pre-fabricated, algorithmically optimised, and emotionally engineered by a system indifferent to the dreamer’s actual psychological needs?

• Hobson and McCarley’s activation-synthesis model (1977): dreams as the brain’s interpretation of internally generated signals, coloured by memory

• Stickgold et al. (2000): Tetris study — subjects who played the game extensively began dreaming in Tetris imagery within days[16]

• Implication: the visual diet of waking life reconfigures the palette of the dreaming mind with measurable, documented speed

Robert Stickgold’s Tetris experiments at Harvard are particularly instructive. Subjects who played Tetris for extended periods began, within three to five days, to report hypnagogic imagery — the visual hallucinations that occur at the threshold of sleep — consisting of falling Tetris blocks. The visual logic of the game had been sufficiently internalised that it colonised the imagery of the pre-dream state.^[16] Now scale this finding to the contemporary situation. A person who has spent six hours consuming short-form video — a diet of algorithmically selected, aesthetically pre-processed, emotionally calibrated visual material — dreams from that archive. They are dreaming, to a degree that should genuinely concern us, from the aesthetic logic of an algorithm.

VII. The Anthropology of Inner Image-Making

Every culture that has maintained oral traditions as its primary mode of knowledge transmission — and anthropology has documented hundreds of such cultures across the breadth of human history — has invested heavily in the cultivation of the inner image. The Aboriginal Australian practice of Songlines, as documented by Bruce Chatwin and elaborated by the anthropologist Deborah Bird Rose, encodes the landscape in sung narrative: the practitioner must construct, from language alone, a mental map of terrain across hundreds of kilometres.^[17] This is a highly sophisticated cognitive technology for maintaining spatial knowledge through imaginative reconstruction — a technology that presupposes, and actively trains, extraordinary capacity for inner picturing.

The Homeric tradition — whatever one’s position on the oral composition debates — produced listeners who could hold the entire topology of the Trojan war, the genealogies of its participants, the geography of the Aegean, and the interior lives of its characters as a sustained imaginative construction across hours of performance. These were active co-constructors of the narrative world, building it image by image in the theatre of the mind. The contemporary equivalent — the viewer of a prestige television adaptation — receives all of that construction pre-assembled, at cinematic scale, with a score designed to prescribe the emotional register at every moment.

The oral tradition trained imagination. The literary tradition trained it differently. The screen tradition, at its current velocity, is doing something no tradition before it has done: it is doing the imagining for us.

Carl Jung, whose theory of the collective unconscious remains contested in its metaphysics but productive in its phenomenology, argued that the imagery of the dream is drawn from two sources: the personal unconscious, constituted from repressed and unprocessed personal experience, and the collective unconscious, constituted from the accumulated symbolic material of the culture into which the individual is born.^[18] Jung could not have anticipated the degree to which the collective symbolic material of the early twenty-first century would be constituted, in substantial part, by the output of machine learning systems trained on the aesthetic preferences of advertising agencies. His framework nonetheless suggests a direction of inquiry worth pursuing with urgency: if the collective visual vocabulary of a culture is increasingly generated by AI, what happens to the symbolic depth of the dreams that draw from it?

VIII. The Argument That Remains

There is a position, fashionable in certain quarters, that holds all of this as merely the latest iteration of perennial panic about new media — that Plato’s anxieties about writing, or the Victorian alarm about novels, or the twentieth-century concern about television, are structurally identical to the present anxiety, and that human cognitive adaptability will find its equilibrium. This position is comforting. On the present evidence, it is also insufficient.

The difference in kind — rather than merely degree — between previous media transitions and the current one is the removal of the imaginative requirement. Reading a novel demands image-construction. Listening to radio drama demands it. Even watching early cinema, with its silent frames and intertitles, demanded it. The contemporary visual media ecosystem, at its highest-engagement edge, has systematically eliminated every point of friction at which the viewer’s imagination might have been required to participate. AI-generated imagery eliminates the last of these frictions: the awareness that a human being, with a specific vision and a specific set of choices, made this image. The AI image is frictionlessly receivable. It asks nothing of you. And the mind that is asked nothing of, over sufficient time, loses the capacity to offer anything.

The neuroscientist Antonio Damasio, whose work on the somatic basis of consciousness and emotion has been among the most consequential contributions to cognitive science of the past three decades, argues that the self is fundamentally an imaginative construction — a continuous narrative that the brain assembles and reassembles from the material of its experience.^[19] The integrity of that narrative depends on the integrity of its raw material. When the raw material of the self’s imaginative construction is, in growing proportion, sourced from outside the self — when the images from which you build your interior world are generated by a machine calibrated to maximise your engagement rather than your growth — the question of who is doing the dreaming becomes, in every sense, an open one.

What do you dream when the world goes silent? The honest answer, for more of us than is comfortable to admit: someone else’s images, in someone else’s palette, assembled by a system that has never once been surprised by a forest.

The forest you imagined at the beginning of this essay — whatever arrived, whatever quality of light, whatever smell or sound preceded the image — that was yours. It was assembled from your actual life: something seen, something read, something felt in the body on a particular morning that lodged itself in memory without your permission. Hold onto it. Because the argument of this essay, if it has made its case, is that the capacity to generate that image — to picture your own forest, dream your own dream, build your own interior world from the material of genuine experience — is a cultivated capacity. And cultivated capacities, without practice, without protection, without the deliberate choice to feed the mind from sources that demand something of it in return, can be surrendered.

The anthropology of dreaming is ultimately an anthropology of the self in its most unguarded state — stripped of performance, of curation, of the managed presentation that waking social life demands. What arrives in that state is what the mind has truly been fed. The question of what we are feeding it — whose images, whose narratives, whose aesthetic logic — is a question about technology only at its surface. At its depth, it is a question about what kind of inner life remains possible when the world goes silent, and whether, in that silence, the voice that speaks is recognisably your own.

---

References & Citations

1. Simpson, J. (2017). Finding Brand Success in the Digital World. Forbes. Updated by Statista Visual Media Exposure Report, 2024.

2. Howes, D. (2003). Sensual Relations: Engaging the Senses in Culture and Social Theory. University of Michigan Press.

3. Potter, M.C., et al. (2014). Detecting meaning in RSVP at 13 ms per picture. Attention, Perception, & Psychophysics, 76(2), 270-279.

4. Levitin, D. (2014). The Organized Mind: Thinking Straight in the Age of Information Overload. Dutton.

5. Adobe / Midjourney usage estimates. Reuters Technology Desk, 2024. Crossref: Stability AI Annual Report, 2024.

6. Berger, J. (1972). Ways of Seeing. BBC & Penguin Books.

7. Wolf, M. (2018). Reader, Come Home: The Reading Brain in a Digital World. HarperCollins.

8. Bureau of Labor Statistics, American Time Use Survey (2024). Crossref: NEA Survey of Public Participation in the Arts, 2022.

9. Piaget, J. (1952). The Origins of Intelligence in Children. International Universities Press.

10. Calvert, S.L., & Valkenburg, P.M. (2013). The influence of television, video games, and the internet on children’s creativity. Creativity Research Journal, 25(1).

11. Turkle, S. (2015). Reclaiming Conversation: The Power of Talk in a Digital Age. Penguin Press.

12. Simmel, G. (1903). The Metropolis and Mental Life. Trans. in The Sociology of Georg Simmel, ed. Wolff, K. (1950). Free Press.

13. Clark, T.J. (2006). The Sight of Death: An Experiment in Art Writing. Yale University Press.

14. Walker, M. (2017). Why We Sleep: Unlocking the Power of Sleep and Dreams. Scribner.

15. Hartmann, E. (1995). Making connections in a safe place: Is dreaming psychotherapy? Dreaming, 5(4), 213-228.

16. Stickgold, R., et al. (2000). Replaying the game: Hypnagogic images in normals and amnesics. Science, 290(5490), 350-353.

17. Chatwin, B. (1987). The Songlines. Jonathan Cape. Extended by Rose, D.B. (1996). Nourishing Terrains. Australian Heritage Commission.

18. Jung, C.G. (1959). The Archetypes and the Collective Unconscious. Collected Works, Vol. 9. Princeton University Press.

19. Damasio, A. (1999). The Feeling of What Happens: Body and Emotion in the Making of Consciousness. Harcourt.