On using distributed computing in searching for the genetic code that enacts energy-negative, carbon-negative photosynthesis

Decomposing Elon’s challenge into parallel stages, we offer the following proposal, which is feasible with current knowledge and material base.

Stage #1 – Physical->Virtual (Entropic->Categorical) Problem Space Mapping

The specification of the problem space should be categorically-covered (i.e. should make use of available knowledge) and entropically-covering (i.e. should represent the modelled process without unaccounted losses of information about states or interaction)

A high-level example of such mapping for the problem of carbon-capture is as such:

<< Find a family of reliable Paths (i.e. graphs of chemical reactions) in the space of known* chemical compounds that convert CO2 into C and other compounds that respects the following:

  • The Paths defined are Hamiltonian over the space of material side effects (i.e. no leftover materials)

Over each Path, define a cost vector (comprised of energy expense, heat exhaust, material intake, real-time duration).


Then, in the defined family of Paths, search for less and less expensive Paths as such.

  • Incentivize some agents to propose new Paths which are likely to have a lower cost vector than existing paths (Innovators)
  • Incentivize some other agents to compute and compare the cost vector of new paths (Accountants)
  • Pay rewards for proposals (attribution) based on the results of the computational

Stage #2 – Computational Distribtuion

The two types of agents (Innovators and Accountants) are under varied computational burdens:

  • Innovator-agents need access to as many results from others’ Proposals as fast as possible (constrained by informational horizon)
  • Accountants need access to fast, collocated, energy-efficient processing power (GPU, FPGA, specialized hardware) to speed-up the “simulation of the proposed reaction” (constrained by computational power)

It follows from this characterization that innovators are many (anyone can come up with an idea), but resources for computing/simulating the validation (and thus accounting for the claim).

Stage #3 – Make sure that Agents are federated and Proposals tradable

Consider defining a vector or a tensor currency in which the proof-of-work is constituted by the distance crowd-measured to the common-objective: energy efficiency carbon capture/carbon negativity.

Stage #4 – Make the mining tools free and point them in the right direction.

It follows to the intuition that the most energy-efficient solution for capturing-carbon is one that needs to depend only on the smallest, most-reliable biochemical factory known to us. The smallest chemical factory is the biological cell. So our lookup/search procedure may be restricted to only those Paths (graphs of chemical reactions) which are self-sustainining (autopoiesic).

As such, the family of Paths that we are looking for is most likely to be found as a life-form (eg. bacteria) which has a metabolism reliant on trapping or converting CO2, but at an energy loss (unlike photosynthesis, which is energy positive).

Stage #5 – Encourage the economy to profit off the new vector currency

Binding the pursuit for the “energy-negative photosynthesis” will allow Innovator agents and Accountant agent to collaboratively-compete with as little of an unnecessary reduction of the solution space as possible.

Summarizing for Geeks

The global challenges we are facing are dependant of our tackling progressive approximation of NP-Complete problems as scales thought unimaginable before, using the new weapons in our arsenal: Deep Learning, Genetic Encoding, Quantum Computing and Ubiqutuous Internet.

We can no longer afford to let the “divide et impere” of big, important problems (i.e. such as finding the energy-efficient carbon-sucking mold, which you can feed and water selectively) to depend on how quickly or efficiently humans can discuss such problems in natural language.

Summarizing for non-geek humans

What I’m proposing in response to Elon’s challenge here is enrolling the super-powers of deep learning and distributed computing to search for the DNA of a new bacteria that will solve the specified problem.



Bogdan Written by: