evolutionary optimization

The main bottleneck in evolutionary algorithms is always the simulation (expressivity, complexity, open-endedness, …).

Evolutionary Optimization In A Nutshell / Disambiguation

AKA. Evolutionary Computation (EC) and Evolutionary Algorithms (EAs) are equivalent, context-dependent umbrella terms.
People say evolutionary optimization when stressing objective-driven improvement, and evolutionary search when emphasizing exploration/diversity (e.g., novelty, map-elites).

In a nutshell. (Generally) zero order black-box optimization methods inspired by biological evolution / natural selection.
The optimization procedure produces a set (population) of candidate solutions (individuals) that evolve over generations.
Population-based methods iterate selection and variation over the population.
Distribution-based methods explicitly maintain and adapt a search distribution ($\mu ,C,\sigma ,\dots$) over the solution/parameter space.

Core families.
genetic algorithm
evolution strategies
differential evolution
genetic programming
evolutionary programming
estimation of distribution algorithms.

Paradigms.
Single-objective
multiobjective
quality–diversity/illumination

Domains on top.
neuroevolution (e.g., NEAT/HyperNEAT)
program synthesis/AutoML
robotics & morphology
etc.

Biggest problem w evolutionary methods (including GA) is memory - you have to store all the weights, and for every member of the population. - Asteriskcat

Is it really a problem tho? Weights/genome should be small. And also you don’t need historic population weights.
Then I see mostly upsides to contemprary methods in that regard actly.