Two very different ideas
Mind uploading, or whole brain emulation, aims to copy the brain itself. The idea is to record every neuron and connection in enough detail, then run that structure as software so that it behaves like the original brain.
An AI clone copies nothing from the brain. It is a language model that has been given examples of how a person writes, plus a searchable store of their messages or notes. It predicts what that person might plausibly say. In other words, uploading starts from biology, while a clone starts from behaviour recorded as text.
The roadmap: what uploading would require
The most cited plan is “Whole Brain Emulation: A Roadmap”, a 2008 technical report by Anders Sandberg and Nick Bostrom at the Future of Humanity Institute, University of Oxford. It was based on a 2007 expert workshop and describes three core capabilities:
- Scanning: imaging a whole brain at a resolution fine enough to see the smallest relevant structures.
- Translation: turning those images into a model, by tracing cells, identifying synapses and estimating their properties.
- Simulation: running the model on hardware fast enough, with a simulated body and environment.
The report lists eleven possible levels of detail, from abstract brain regions down to individual molecules. Workshop participants tended to think that somewhere between a spiking neural network and a model including chemical concentrations would be needed, but nobody knows for certain which level is enough.
Real connectome milestones
A connectome is a complete map of the connections between nerve cells. It is the first ingredient that uploading would need. These are the landmark results so far:
| Year | Organism and sample | What was mapped | Published in |
|---|---|---|---|
| 1986 | Nematode worm C. elegans | All 302 neurons of the hermaphrodite and their connections | Philosophical Transactions of the Royal Society B |
| 2024 | Human cortex, about one cubic millimetre (H01) | About 57,000 cells and 150 million synapses; 1.4 petabytes of image data | Science |
| 2024 | Adult fruit fly, whole brain (FlyWire) | 139,255 neurons and about 50 million synapses | Nature |
| 2025 | Mouse visual cortex, about one cubic millimetre (MICrONS) | More than 200,000 cells and over 500 million synapses, linked to recorded activity | Nature |
Progress is real and fast. The fly map relied on artificial intelligence to trace cells, then on human proofreaders and citizen scientists to correct it. In the same 2024 publication package, a team led by researchers at UC Berkeley ran a simplified model of the whole fly brain on a laptop and used it to predict which neurons would respond to certain stimuli.
The gaps that remain
- Scale. The 2008 roadmap counted roughly 1011 neurons and 1015 connections in a human brain. The Harvard team behind H01 said that mapping a whole mouse brain would need about a thousand times the data of their one cubic millimetre, and a human brain is far larger again.
- Wiring is not activity. A connectome shows which cells connect, not everything about how strongly they signal, how chemicals change them or how they learn. The fly model was a useful simplification, not a living fly mind.
- Scanning is destructive. Today’s nanometre-scale maps come from preserved tissue cut into very thin slices. There is no method for scanning a living human brain at this resolution.
- Computing is not the main barrier. The roadmap estimated about 1018 operations per second for a spiking-network emulation, and the fastest supercomputers now exceed that. The harder problems are scanning, interpretation and knowing which level of detail matters.
- Identity is unresolved. Even a flawless emulation raises the question of whether it would be you or a new being that shares your memories.
The Nectome controversy
In 2018 a start-up called Nectome promoted a brain preservation method intended to keep the connectome intact for possible future uploading. MIT Technology Review reported that the procedure would be fatal for the person and was aimed at terminally ill people. The company had already taken deposits from people hoping to have their brains preserved.
Following the coverage, the MIT Media Lab ended a research subcontract with the company, citing the scientific premises behind its commercial plans. One neuroscientist quoted by the magazine called the company’s core proposition “just false”. The episode shows how easily the promise of uploading can run ahead of the science.
How AI clones work, by contrast
An AI clone exists today because it asks for far less. It uses a general language model, examples of a person’s writing and a search index of their messages. Our guide to how AI clone memory works explains the retrieval step, and AI clone versus chatbot explains what makes it personal.
| Question | Mind uploading | AI clone |
|---|---|---|
| What is copied? | The brain’s structure and dynamics | Patterns in a person’s words |
| Input | A complete, nanometre-scale brain scan | Messages, notes or recordings |
| Exists today? | No | Yes |
| Has its own experiences? | Unknown and debated | No |
| Main risk | Unproven science, irreversible procedures | Invented replies, privacy and consent |
| Honest description | A research goal | An AI simulation of how someone wrote |
Neither option offers a way to live on. Our article on whether you can live forever with AI looks at medical research and big predictions separately.
Where Memory Clone fits, honestly
Memory Clone is an AI clone tool, not a step towards mind uploading. It is a Windows app in alpha that builds a private AI simulation from chats you export yourself and notes you write. It learns writing style separately for each relationship and keeps imported originals on your PC. It does not use voice, photos or video today, and every clone is presented as an AI simulation, not the real person behind it. You can read more on our AI transparency page or the download page.
Frequently asked questions
Has anyone uploaded a mind yet?
No. Scientists have mapped the wiring of a whole fruit fly brain and small pieces of mouse and human brain, but no brain of any animal has been fully emulated.
When will mind uploading be possible?
Nobody knows. It depends on scanning, interpretation and modelling problems that are not yet solved, so any date is a guess rather than a forecast.
Is an AI clone a first step towards uploading?
Not really. A clone learns from text and has no access to brain structure, so improving clones does not bring uploading closer.
Would an uploaded mind be me?
Philosophers disagree. Some think a sufficiently accurate copy would continue you, others think it would be a new person with your memories.
What is a connectome?
A complete map of the connections between nerve cells in a brain or part of a brain.
Sources
Checked on . Third-party products and help pages change often; follow the links for current details.
- Sandberg & Bostrom, Whole Brain Emulation: A Roadmap (Technical Report 2008-3) (Future of Humanity Institute, University of Oxford)
- White, Southgate, Thomson & Brenner, “The structure of the nervous system of the nematode Caenorhabditis elegans” (1986) (Philosophical Transactions of the Royal Society B)
- Dorkenwald et al., “Neuronal wiring diagram of an adult brain” (2024) (Nature)
- Shiu et al., “A Drosophila computational brain model reveals sensorimotor processing” (2024) (Nature)
- MICrONS Consortium, “Functional connectomics spanning multiple areas of mouse visual cortex” (2025) (Nature)
- Shapson-Coe et al., “A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution” (2024) (Science)
- The brain as we’ve never seen it (2024) (Harvard Gazette)
- Mapping the entire fly brain (2024) (Princeton Engineering)
- MIT severs ties to company promoting fatal brain uploading (2018) (MIT Technology Review)
- TOP500 supercomputer lists (TOP500)
This article is general information, not medical, legal or financial advice. Memory Clone creates an AI simulation from the material you provide; it is not the person.