Short: Collect brain data.

Long: After a patient’s brain has been cryogenically frozen to where all synapse states are preserved, the brain is sliced into thin segments. These slices are scanned and the images from the scans undergo analysis of the structure of the brain tissue. With recognition of structures in the brain, and their relations to the surrounding structures, the functionality of the brain tissue scanned is interpreted and recorded as data.

Short: Analyze structure and activity data.

Long: System identification for each neural circuit depends on a tremendous amount of recorded activity data and, to a lesser extent, on more generic connectome structure data.

short: Initialize a generic brain model

long: To begin, we initialize a feasibly-derivable, constrained model with generic knowledge of circuit connectome and suitable kernel functions.

short: Upgrade

long: Prosthetic devices for brain regions could one day enable the possibility of upgrades, potentially helping with mental illness or even new mental capabilities.

short: Update

long: Just as a patient with a cochlear implant hearing aid today can have their hearing aid improved as updates are developed, so too could neural prosthetics be improved by future updates.

short: Place neural prosthetics in brain regions

long: In this phase, a brain region is replaced by a neural prosthetic which implements the validated model of that brain region. The neural prosthetic emulates the replaced brain region and provides all of the same activity and functionality.

short: Implement model – neural prosthetics

long: Once the model is fine-tuned and validated, it can be implemented in the form of prosthetic devices which replace the modeled regions of the brain.

short: Validate the model

long: We continue fine-tuning the model, introducing controlled input variables corresponding to real-world stimuli (e. g. perception of a color or experience of an emotion), until there are no discrepancies and its output matches the original data.

short: Fine-tuning the model

long: In this phase we use the model to emulate the original brain tissue, recreating its neural activity. The outputs of the emulations are compared against the original brain data, and the parameters of the emulation are adjusted.

Implement the Model – Neural Prosthetics
It might eventually be possible to achieve mind uploading by using neural prosthetics to replace specific brain regions one at a time. This approach is known as “gradual replacement.” If the gradual replacement approach is possible, a patient could have one or more regions of their brain replaced with prosthetics.
Once the model is fine-tuned and validated, it can be implemented in the form of prosthetic devices which replace the modeled regions of the brain.

Place Neural Prosthetics in Brain Regions
In this phase, a brain region is replaced by a neural prosthetic which implements the validated model of that brain region. The neural prosthetic emulates the replaced brain region and provides all of the same activity and functionality.

Fine-Tuning the Model
In this phase we use the model to emulate the original brain tissue, recreating its neural activity. The outputs of the emulations are compared against the original brain data, and the parameters of the emulation are adjusted.

Validating the Model
We continue fine-tuning the model, introducing controlled input variables corresponding to real-world stimuli (e. g. perception of a color or experience of an emotion), until there are no discrepancies and its output matches the original data.

Collect Brain Data
After a patient’s brain has been cryogenically frozen to where all synapse states are preserved, the brain is sliced into thin segments. These slices are scanned and the images from the scans undergo analysis of the structure of the brain tissue. With recognition of structures in the brain, and their relations to the surrounding structures, the functionality of the brain tissue scanned is interpreted and recorded as data.

Extract Neuronal Activities
System identification for each neural circuit depends on a tremendous amount of recorded activity data and, to a lesser extent, on more generic connectome structure data.

Initialize the Model
To begin, we initialize a feasibly-derivable, constrained model with generic knowledge of circuit connectome and suitable kernel functions.

Update
Just as a patient with a cochlear implant hearing aid today can have their hearing aid improved as updates are developed, so too could neural prosthetics be improved by future updates.

Upgrade
Prosthetic devices for brain regions could one day enable the possibility of upgrades, potentially helping with mental illness or even new mental capabilities.