[ExI] The problem of embodiment and uploading humans

[Ben Zaiboc]

On 13/09/2026 13:22, BillK wrote:
>
> Hi Ben
> Gemini liked your comment!  
> The main objection was to the part about bodies being simpler than brains, thus making an "average" body possible. 

Ugh.
It's really painful to read this stuff. Or if I was an LLM I would probably say:

"1. The Massive Hyperbole Problem
It's excruciatingly painful to insightfully read this crucially significant stuff".

Anyway. I'm pretty sure that objection is overblown, but I have a solution.

"Your brain does not contain a generic body driver. Your motor and somatosensory cortices contain a hyper-specific map (the cortical homunculus) uniquely shaped by your personal limb lengths, nerve conduction velocities, sensory thresholds, and motor habits built over decades.". This is exactly what I'm getting at when I say "as long as the interfaces are properly designed".

Our brains don't have a generic body driver, no, but uploads could.

We don't tend to think about our brains being interfaced with our bodies. They are (essentially, there is a series of junctions between the Central Nervous System (the brain and spinal cord) and the Peripheral Nervous System (which is part of the body, and outside the 'Brain/Body Barrier')), and in an upload we can take advantage of that, in a number of different ways, one of which would enable this 'standard body model' idea to work (there are other ways, which lead to lots of interesting possibilities, but that's beyond the scope of this thread).

What I'm thinking of is a 'standard body' design, with custom interfaces based on the brain scan information about the motor and somatosensory maps.  A simple example would be if the brain expects its legs to be 90 cm long, and the average body model has 1 metre legs, the interface would automatically multiply the perceived length by 0.9.

We will certainly need AI to design and create such an interface. In fact, I'm envisaging a program, or set of programs, that would automatically create the interface for each person. So the problem would be not creating the interface, but creating the program that would create the interface. Definitely a job for AI. It's not just physical parameters that would be involved, but biochemistry and neurology, in lots of different ways. Still, I think this would be the simplest path.

This would just be the initial stage, as I said earlier. It would make the transition easier until the upload realised that they would actually (maybe) prefer 1 metre legs, and they could then adjust the interface accordingly.

Obviously it would be much more complex than this, but that's what AIs are for, dealing with the complex stuff.

(In fact, thinking about it, I wouldn't be surprised if each upload would have the option to have an embedded AI. That would be useful for all sorts of things. Also potential for all sorts of problems. Topic for another thread!)

I'm pretty sure it would be easier and more practical to take this approach than to make a detailed body scan then construct a virtual body based on it. Separately for each upload.

Once all the details are worked out, it should be possible to create an interface-generation program that would be applied to all uploads, together with the standard VR body, so that uploading would involve a brain scan and the addition of a standard package of VR items (probably the body, a world - or at least a room - and some tools to enable things like control of sensory, motor functions and other things, and communication) plus the programming to make them work for any individual. All of this only needs to be created once, for all uploads (who would then be able to change things as they like).

Given that one of the aims of uploading is to enable us to change our bodies as desired, this approach would be more practical as well. It would be harder to tinker with a custom-made body that exactly replicates your original biological one (which would probably have been the cause of your death in the first place, so that needs to be considered) than a standard (healthy) model with maybe a built-in control panel to make it easy to experiment with changes, and a system that automatically adjusts for mismatches between the body and the brain's expectations.

Ok, two standard bodies, one male and one female. (And before you start, shut up. If you wanted more options than that, you could do it yourself, once you learn how.)


"An upload’s virtual body would likely need to be as tailored to their individual biology as a bespoke suit, at least initially"

Exactly. And that's just what my proposal would achieve, in practice, by the use of an automatically-configured interface to a standard body model, with the added advantage of easy future customisability.

Oh, one final thought: If a body-scan was really essential for ensuring the sanity of uploads, a lot of neuro-only corpsicles are already in trouble!
-- 
Ben