When will the first human whole-brain emulation, capable of being called a true 'mind upload', be created from a deceased human's preserved and imaged brain?

As of mid-2026, whole-brain emulation (WBE) remains vastly far from the resolution criteria. The current state of the art in connectomics includes the complete adult fruit-fly connectome (~130k neurons) nimh.nih.gov and millimeter-scale mammalian reconstructions, such as a 1 mm³ mouse cortex nature.com and a 1 mm³ human cortex sample (~57,000 cells, ~150 million synapses) pmc.ncbi.nlm.nih.gov, [PMC11718559]. Scaling from this 1 mm³ sample to an entire human brain (~86 billion neurons, >100 trillion synapses ninds.nih.gov) represents a million-fold leap in volumetric imaging, data storage, and automated reconstruction.

While preservation techniques are relatively advanced and not the primary bottleneck—demonstrated by aldehyde-stabilized cryopreservation for large mammals brainpreservation.org and recent protocols compatible with physician-assisted death biorxiv.org—the transition from a static preserved structural map to a functioning cognitive model is an immense hurdle. Current functional models, even at the Drosophila scale, explicitly omit critical variables such as morphology, gap junctions, internal molecular states, neuropeptides, and receptor dynamics .

This specific forecast is heavily constrained by its strict criteria: producing a 'true mind upload' from a deceased person that achieves broad expert consensus. Deriving dynamic functional weights and molecular states from an aldehyde-preserved brain without a living reference to validate against poses profound scientific and philosophical challenges. Baseline community forecasts (such as a Metaculus median of ~2068 metaculus.com) and the 2025 State of Brain Emulation report (projecting a 30-40 year lower bound mindtransfer.me) generally target generic emulation milestones. Because this question demands post-mortem validation and expert consensus on personal cognition and identity, the timeline must be extended further.

The resulting distribution balances potential AI-driven acceleration with severe functional-validation bottlenecks:

• Left Tail (P10: 2054, P25: 2068): Assumes transformative AI radically compresses the engineering timescales for molecular parameter inference, automated segmentation, and exascale imaging infrastructure lesswrong.com. The P10 represents a scenario bounded primarily by hard physical constraints, such as raw scanning throughput and mandatory biological validation cycles. • Median (P50: 2088): Centers late in the 21st century. This allows multi-decade timelines to solve the physical exascale imaging problem, followed by the harder neuroscience challenge of molecular/functional translation, and the inherently slow process of building scientific consensus that a computational simulation truly replicates a deceased person's original cognition. • Right Tail (P75: 2135, P90: 2230): Maintains a very long, fat right tail to account for the risk that static connectomes irreversibly lose critical memory-relevant state variables, the possibility that deriving dynamic weights from static preservation requires fundamentally new paradigms in physics or biology, and the historical trend of over-optimism in transhumanist timelines.