There’s no guarantee that an evolutionary search process will lead to a globally optimal solution. It’s the same thing with evolutionary algorithms in computing
The problem is that the landscape of where the global maxima are changes faster than evolution can keep up. If the environment were entirely static, then yes, mathematically speaking any random optimizer would eventually reach a global maximum. However, it could take, say, 1050 years or more to jump from a local maximum to a distant, higher maximum.
Imagine an alligator. Quite good at catching prey with their current anatomy.
An alligator that shoots laser beams for tracking and bullets would be even better. There’s however no path from their current anatomy to this state, regardless of the randomness and timescale for mutations. In fact, in order to achieve this higher state several non advantageous intermediates would be necessary and therefore never selected for.
So no, evolution can’t achieve global maxima, it can however optimize the shit out of what it’s given to work with.
There needs to be pressure for animals with a mutation to reproduce more or animals without the mutation to die before reproduction. Like a disease for example. Otherwise the genes don’t spread and just disappear in the soup of all that species genes and never become dominant. Without any evolutionary pressure the mutation will only spread in one family and probably be gone after a few generations. Like there are human families that are more likely to produce offspring with 6 digits on their hands, but since it isn’t more advantageous than 5 digits (6 digit people don’t produce more offspring and 5 digit people aren’t more likely to die before reaching reproductive maturity) that mutation doesn’t spread across the entire species. Sure if you could sample the genome of every human on earth and identify every advantageous gene mutation you could build the ultimate human DNA. But that’s artificially created, something like that will never happen through evolutionary pathways.
Dont they eventually produce global maxima by iterating towards it through the many degrees of freedom allowed by crazy mutations and time?
There’s no guarantee that an evolutionary search process will lead to a globally optimal solution. It’s the same thing with evolutionary algorithms in computing
The problem is that the landscape of where the global maxima are changes faster than evolution can keep up. If the environment were entirely static, then yes, mathematically speaking any random optimizer would eventually reach a global maximum. However, it could take, say, 1050 years or more to jump from a local maximum to a distant, higher maximum.
Ah good point, thanks
Imagine an alligator. Quite good at catching prey with their current anatomy.
An alligator that shoots laser beams for tracking and bullets would be even better. There’s however no path from their current anatomy to this state, regardless of the randomness and timescale for mutations. In fact, in order to achieve this higher state several non advantageous intermediates would be necessary and therefore never selected for.
So no, evolution can’t achieve global maxima, it can however optimize the shit out of what it’s given to work with.
There needs to be pressure for animals with a mutation to reproduce more or animals without the mutation to die before reproduction. Like a disease for example. Otherwise the genes don’t spread and just disappear in the soup of all that species genes and never become dominant. Without any evolutionary pressure the mutation will only spread in one family and probably be gone after a few generations. Like there are human families that are more likely to produce offspring with 6 digits on their hands, but since it isn’t more advantageous than 5 digits (6 digit people don’t produce more offspring and 5 digit people aren’t more likely to die before reaching reproductive maturity) that mutation doesn’t spread across the entire species. Sure if you could sample the genome of every human on earth and identify every advantageous gene mutation you could build the ultimate human DNA. But that’s artificially created, something like that will never happen through evolutionary pathways.
No.