|
Getting your Trinity Audio player ready…
|
Memory Outside the Head?
For more than a century psychologists and neuroscientists have treated memory as something the brain stores the way a hard drive stores files: you experience an event, neurons change, the traces sit there until you recall them. Rupert Sheldrake turned that picture upside-down in the 1980s when he proposed that living things do not keep full “copies” of their past inside their cells at all. Instead, he said, every organism taps into a vast, invisible informational field—what he calls a morphic field—and the field itself carries the memories of similar events across time and space. In Sheldrake’s words, the brain would behave less like a tape recorder and more like a radio tuning in to broadcasts that already exist in the ether. (sheldrake.org)
How the Morphic Field Is Supposed to Work
Imagine you are the first person ever to solve a Rubik’s Cube. According to morphic-field theory, your success slightly strengthens a pattern in nature. The next solver, perhaps continents away, will unconsciously resonate with that pattern and finish the cube a little faster, and each repetition reinforces the field further. Sheldrake applies the same logic to the way termites build mounds, the way a fetus organizes itself, even the way fundamental laws of physics might “solidify” as the universe ages. The key mechanism is similarity: present actions “lock onto” past ones that look alike, the way a singer’s voice can set off vibrations in a wine glass of just the right pitch. (sciencedirect.com)
Anecdotes, Rats, and Running Mazes
Sheldrake’s favorite laboratory tale involves generations of white rats learning a water-maze in London. He reported that once many rats had mastered the trick, completely new litters on the other side of the Atlantic supposedly learned it faster too—no contact, no shared genes, just a worldwide “habit” forming in the field. He collects similar stories of people who have sudden flashes of insight and pets that seem to know when their owners decide to come home. All, he says, point to a transmissible memory that sits outside any single brain. (sheldrake.org)
Why Most Scientists Remain Unconvinced
Mainstream biology replies that we already have rock-solid mechanisms—DNA, synaptic plasticity, cultural teaching—to explain those results without invoking an invisible field. Controlled replications of the maze story have failed, and attempts to measure morphic effects directly have not beaten statistical chance. The consensus labels the idea “pseudoscience” until it yields a testable equation and passes reproducible experiments. (en.wikipedia.org)
A Surprising Echo in Artificial Intelligence
Yet a curious parallel has emerged in modern artificial-intelligence research. Large-language models (LLMs) such as GPT-4o or Claude began life as gigantic pattern recognizers whose internal parameters encoded nearly everything they knew. But as the internet kept growing, retraining those models from scratch became too costly and too slow. Engineers responded with retrieval-augmented generation (RAG), a design that splits “thinking” from “remembering.” The neural net now asks an external database for relevant passages and then weaves an answer in real time. (arxiv.org)
How External Memory Works in Plain English
Picture a diligent student with a photographic mind but no bookshelf. Traditional LLMs tried to memorize the entire library before the exam and then relied only on memory during the test. A RAG system, by contrast, keeps a slim, nimble brain and a giant library card. When you pose a question—“Who signed the Kyoto Protocol?”—the model converts your sentence into a mathematical fingerprint, looks up fingerprints that sit nearby in the vector catalog, pulls down the matching paragraphs, and only then starts writing its answer. If the catalog gets new climate-policy documents tomorrow, the model benefits instantly without any surgical brain operation. (promptingguide.ai)
Why Companies Love the Idea
Businesses quickly noticed that RAG lets them bolt a private knowledge base onto a public LLM. Workday pipes internal HR manuals into a vector index so employees can ask policy questions; Bloomberg does the same with market data. Because the facts live outside the model, they can be audited, updated, or deleted on demand—a crucial feature in regulated industries. (wsj.com, time.com)
The Transformer as a “Receiver”
Once a retrieval round finishes, the transformer network behaves a bit like a radio pulling together multiple stations: it lines up each retrieved paragraph next to your question and pays selective attention across the whole bundle. The answer you read is thus a live composition that depends on traffic between the weights (the brain) and the documents (the field).
Where the Two Stories Meet
Sheldrake argues that memory is not trapped inside the skull; RAG engineers deliberately move factual load outside the silicon “skull” of their models. Both frameworks rely on similarity for access—morphic resonance in one case, vector-space nearest neighbors in the other. Both predict that repeated events leave a stronger trace, making future recall easier. And both cast the local computing unit—brain or neural net—as a receiver or compositor rather than an all-in-one warehouse.
Why the Resemblance Is Only Skin-Deep
Look more closely, though, and the kinship fades. RAG runs on solid-state drives you can point to; morphic fields lack a known physical carrier. Vector search uses cosine distances you can write on a chalkboard; resonance lacks a settled equation. You can pull one PDF out of a RAG index and the model will stop citing it instantly; nobody knows how to surgically remove a single maze-running memory from the cosmic field.
What Brains Actually Reveal Under the Microscope
Neuroscience over the past decade has gone further than ever in tracing memories inside living brains. With optogenetics, researchers tag the exact neurons (“engrams”) that lit up during a mouse’s fear lesson and later reactivate only those cells to replay the memory. Cryo-electron microscopy shows that learning literally reshapes dendritic spines—the tiny knobs where neurons meet—within hours. Blocking the molecular machinery that strengthens those spines erases recall. Those findings leave little doubt that at least some memory lives in tangible tissue. (nature.com, nature.com, biorxiv.org)
Yet Brains Still Behave Like Predictive Receivers
None of that rules out a “receiver” role. Predictive-processing theories describe the cortex as a gadget that constantly guesses what its senses will report next and then corrects its mistakes. The traffic looks uncannily like the query-and-retrieve loop in RAG: the brain issues predictions, internal or external signals answer back, and the net of neurons updates its picture of reality on the fly.
Edge Devices on a Wider Fabric
Seen this way, both biological and artificial intelligences resemble smartphones on a cell network. Each device does plenty of local computing, but value blossoms when they connect to a shared infrastructure—cloud servers in the AI story, perhaps a morphic field in Sheldrake’s. If such a field existed, the brain would merely add another layer to the concentric rings of memory: synapses for personal experience, culture for communal knowledge, morphic resonance for species-wide habits.
Testing the Wild Idea with Modern Tools
Could AI help hunt for morphic effects? One hypothetical experiment: train a robot arm to solve a new puzzle and log the motion trajectories as vectors. Release the data only after labs on three continents try to teach their own arms the same trick. If those distant arms improve at an anomalous rate before the data drop, engineers will have to look for a non-ordinary information path. No result so far hints at such leakage, but the tools to run global, tightly sealed tests now exist.
Practical Wisdom from the Comparison
For AI designers, morphic resonance is a reminder that clever systems might benefit from shared, evolving memory pools rather than ever-bigger private brains. For scientists skeptical of morphic fields, the RAG story shows how to convert a romantic metaphor into a measurable pipeline: specify the storage medium, define the retrieval rule, measure the latency, count the errors.
Remaining Hurdles
Extraordinary claims demand extraordinary evidence. In morphic-field research the bar is still unmet. On the AI side, external memory introduces fresh vulnerabilities: broken links, poisoned documents, privacy leaks. Both visions thus face the same bottom-line question—how do you trust what comes in over the airwaves?
A Middle Path: Humility and Experiment
Perhaps the safest stance is open-minded humility. Nature has repeatedly surprised us with phenomena once deemed impossible—continental drift, quantum entanglement, gut microbes altering mood. The responsible response is neither automatic belief nor reflex dismissal but iterative experiment. If morphic resonance is real, disciplined trials will flush it out. If it is not, tightening the net will refine our understanding of how memory really works, whether in carbon or silicon.
Conclusion
Sheldrake invites us to picture a universe where every action leaves an echo that future minds can tap, while LLM engineers build practical systems that already externalize knowledge in plain sight. The resemblance is compelling: both stories say memory need not live where processing happens, both lean on similarity to pull the past into the present, and both promise efficiency by sharing the heavy lifting. The gulf is just as real: one is metaphysics awaiting a mechanism; the other is a mechanism awaiting continual refinement. By holding the two ideas side by side we sharpen our questions: Where exactly is memory stored? How is it retrieved? What counts as evidence? The answers will illuminate not only how thinking machines grow smarter but also how our own minds, perched between synapses and sky, weave the fabric of experience.
Leave a Reply