Most EAs may be divided into generational algorithms, which update the entire population once per iteration, and steady-state algorithms, which update the sample a few candidate solutions at a time.
Steady-state GAs (popularized by the Darrell Whitley and Joan Kauth's GENITOR system) iteratively breed a new child or two, assess their fitness and then reintroduce them directly into the population, thus killing some preexisting individuals to make room (take a look at Essential Metaheuristics by Sean Luke for further details).
So a steady-state GA fits part of the description ("death comes after random period of time").
Other characteristics ("the better phenotype is (fitness function) the bigger is chance to have (more) children") are shared with generational algorithms.
Gender-specific selection / crossover operators / GAs fit the last part of the description ("phenotypes have sex (man or woman)"):
instead of selecting individuals from a single group for mating, a gender genetic algorithm create two groups and permits mating only between opposite groups; the selection from two groups helps to increase diversity.
Anyway many GA variants based on a more natural governing framework are possible:
- delayed reproduction / breeding age (most life forms aren't able to breed during the early stages of their life);
- life span for individuals (no matter how poor the fitness is, it
would still stay in the population for a length of time) / predetermined death age;
- progressive ageing (Individual Aging in Genetic Algorithms by Ashish
Ghosh, Shigeyoshi Tsutsui and Hideo Tanaka);
- Age-Layered Population Structure (ALPS) evolutionary algorithms (The Age-Layered Population Structure by Gregory S. Hornby)