[Zero]

Well this one is almost useless, but it's so cute I had to share it.

Code

class NaturalSelector {

    constructor(score, mutate) {

        this.score = score;
        this.mutate = mutate;
    }

    evolve(value) {

        let candidate1 = this.mutate(value),
            candidate2 = this.mutate(value);
        
        return this.score(candidate1) > this.score(candidate2) ? candidate1 : candidate2;
    }
}

Usage:

Code: text

let sel = new NaturalSelector(x => x, x => x + Math.random() - Math.random());

let n = 0;

for (let i = 0; i < 10; i++) {
    n = sel.evolve(n);
    console.log(n);
}

[Zero]

If you have requests for my tinytool series, don't hesitate!

[MagnusWootton]

If you hooked that up to a tank gun thats shooting across a mountain to hit another tank, the score could be how close the bullet landed to it.

[HS]

From my point of view, natural selection seems to be an important piece of the AGI puzzle. The trouble is, I think it's too general for what I'd call general intelligence. So I'm wondering if it's possible to write a natural selection program that not only improves the fitness of the system, but also improves the fitness of the natural selection process itself. That way it wouldn't be blind selection, i.e. you could theoretically narrow the search to a pre-set definition of intelligence.

[ivan.moony]

From my point of view, natural selection seems to be an important piece of the AGI puzzle. The trouble is, I think it's too general for what I'd call general intelligence. So I'm wondering if it's possible to write a natural selection program that not only improves the fitness of the system, but also improves the fitness of the natural selection process itself. That way it wouldn't be blind selection, i.e. you could theoretically narrow the search to a pre-set definition of intelligence.

Gödel machine