Immortality: Critical path analysis

“Personally, I’ve been hearing all my life about the Serious Philosophical Issues posed by life extension, and my attitude has always been that I’m willing to grapple with those issues for as many centuries as it takes.”

– Patrick Nielsen Hayden

The problem:

“Because the Reaper comes for all eventually, humans have formed a relationship with death perverse. Like a hostage who grows to love their kidnapper, humans tell themselves the handful of decades the reaper gives them is just the right length, that living a truly long and healthy life would get boring. [..] “…Death is a part of life, he whispers. “Death gives life meaning.”

This is madness.

Misery doesn’t give happiness meaning. Happiness is meaning itself. If you tortured people to make them better appreciate the pleasures of life, you would be a monster. Just like death. No parent would ask the Reaper of Age to wrinkle their child’s skin, weaken their bones, dim their vision, and their minds, cripple them in a thousand ways over decades to ultimately kill them, “to give their life meaning”. “

CGP Grey: Why Die

With everything, that’s important in life, let’s take a diversified portfolio approach. Below are the comprehensive list of strategies for individual long-term survival, along with their respective critical paths: things on which we can start working right now in order to get there in a timely basis:

Attacking mortality via Bioengineering

Executive overview: reaching actuarial escape velocity by slowing down/halting the biological decay with continuously repairing, or rendering harmless the damaged structures. This is essentially the approach pioneered by Aubrey de Grey / sens research foundation.

Critical path: while reaching accuraial escape velocity can happen in a staged approach -ie fix first bottleneck, live to see second bottleneck eliminated, etc- the entire roadmap is daunting. Specifically, SENS recognizes that there are seven major classes of cellular, and molecular damage contributing to aging: Cell loss&tissue atrophy, Cancerous cells, Mitochondrial mutations, Death-resistant cells, Extracellular matrix stiffening, Extracellular aggregates, and Intracellular aggregates. The upside of this approach, is that research proceeds across all of these in parallel. The downside is, that your particular survival rate is determined by the earliest cause, which, at your point of death, have not been fixed yet.

(see also Gwern’s “ A History of Life-Extensionism in the Twentieth Century”: there are no simple interventions that can change average life expectancy by more than a few years or maximum life span at all )

Attacking mortality via uploading & Whole Brain Emulation

The general idea here, is that bodies & brains are transient, running the same process on a less transient hardware -eg computers- would not only allow for immortality, but kickstart the process of rapid recursive self-improvement.

Primary literature for this is the Whole brain emulation roadmap

Critical paths:

  • understanding the level of simulation required for functional mind transfer
  • mapping the required level: Spiking neural network / Electrophysiology / Metabolome
  • running it in a simulation

Assuming a spiking neural-network level resolution for cognition implies 10^18 Flops ~= exaflop-level compute; as of 2018, this is already present. This leaves the biggest problem: mapping an existing brain at the required level for simulation.

Specific ways to do that might include:

Cognitive metamorphosis: Switching neurons one-by-one (AKA Moravec Transfer)

Executive overview: slow replacement of natural neurons into less transient artifical ones, essentially resulting in an upload. Slow transition from evolutionarily engineered neurons to artificial neurons.

Argument for Moravec transfer includes the possibility of staged transfer: Port 1% of the brain, go away for a week, try how it feels like, and check for behavior / cognitive changes; proceed if none. This checks for compatibility of artificial neurons against biological ones.

Brain scanning

Specific approaches include: openWater’s non-intrusive scanning via light, and neuralink neuro lace. The general idea is to make a high-resolution scan of the neurons & connectivity graph in a given brain, then run a real-time simulation on it.

Imitation learning

Understanding of the simulation might be partially alleviate by a hybrid deep learning approach, approximating (measurable) neural network spikes (imitation learning).

This technique requires orders-of-mag lower scanning fidelity, and thereby the closest candidate to reaching human-level (and human-like AI); but might or might not cross the threshold for “digital immortality” for personal aspirations.

Economic consequences of mind uploading: Uploads economics 101

Approaches generally less considered:

  • Mass Cybernatization: Replacing parts of the body as they break down (hearth, lung, limbs, etc), substituting it with ones that don’t (superhuman prosthetics). Artificial limbs have decent traction, artificial organs do not.
  • Head-in-the-jar: the idea here is that people, generally, die by their body giving up -the brain alone has a significantly longer possible lifetime, but organ failure takes it all. Removing the head, and hooking it up to artificially synthesized body replacement might sidestep this. The issue here is generally two-fold: the stomach hosts a nervous system of it’s own (roughly equaling the size of a cat’s brain) which might, or might not factor into cognition; and the replacement involves cellular-level generation. Neither of these are currently pursued by any research I know of.

Failback plans

Cryonics

Preservation after death; good literature is eg Facts about brain presevation. In general, this passes down the buck to future generations.

Beta uploads

Executive overview: “if we could make a large number of observations of a human and then use these to create a plausible computer model of that person’s mind, and if the detail with which the person was observed was high enough, the computer model created in this way may be close enough to the real person’s mind that we may actually consider his/her identity to have been copied into the computer model to a significant degree”

Observations, in sorting order of desirability:

  • DNA samples & healthcare info, neurocortex maps
  • Lifelogging: camera, bio-and neuro-feedback sample gathering
Joel Solymosi @joelsolymosi