Mind · Strategy
ConcurrentBetHedge
Switch one line of development back and forth between two inversely related niches, so that a path to complex function exists where no path exists within a single niche
Definition
Mechanism
In a complex niche, valid functions are separated by gaps: from a working gene, every single added change breaks it before a longer working gene is reached, so no step-by-step path of improvements exists against that one target. Pair it with a simpler niche in which the broken intermediate is an advantage (a disabled pathway lowers metabolic load, as yeast mutations harmful in the wild are beneficial in a lab medium). Then, generation by generation, some fraction of the population finds itself in whichever niche currently rewards its form: the intermediate steps pay off in the simple niche, and the completed longer gene pays off in the complex one. The two niches act like a hedge, being inversely related, but used concurrently over time rather than in parallel. This raises MetaAccuracy at the cost of meta-precision, as a NaturalProblem requires. It differs from Exaptation, which is a single, decisive switch of target, and it can be read as the population's ComputationalEntropyRate being lowered by having two acceptor languages. Done asynchronously over the same code, it can produce overlapping coding sequences.
Preconditions
- Two niches are available whose fitness criteria are approximately inverse
- Some fraction of each generation can occupy either niche, by agency or by distribution
Invariants
- At each step the current form is beneficial in at least one of the niches
- The niches are inversely related: what breaks function in one helps in the other
- The switching is repeated, not a one-off transition
Risks
Failure modes
Evaluating progress only in the complex niche and concluding no path exists
Treating it as a static portfolio split across niches
Assuming the niches are independent, which lowers the odds that each improves
Relationships
Builds on
References
Infrastructure
Bits submitted per bit accepted in a channel that computes, where rejection is meaningful work rather than noise
Physics
A mechanism that evolved for one function is taken over for a different function in a new context
Success at hitting any one of several available targets, each measured in its own code
A problem in which meta-accuracy trades off against meta-precision, so that solving it requires changing shape
Used by
References
Fill the room a prediction frees up with an extra process interleaved with the main one, so a second function runs at no visible cost