But wouldn't it be too late to evaluate the fitness of something that has already become Skynet? Also, what if becoming Skynet is the least expensive option in the problem domain? That's only the equivalent of 22 million water crossings.
You just put in a function that checks if your algorithm is going to become Skynet in the next generation, before you actually apply the fitness function.
That's what I do.
And you don't want the evolution of Skynet to be impossible, just costly.
Nevermind. Spent 10 seconds to look it up myself. Does this mean I lose nerd-cred, both for not automatically knowing this (sorry, watched the first movie, liked it, thought the second was a lot of lame, and never watched anything else dealing with the Terminator), and for not googling it?
But wouldn't it be too late to evaluate the fitness of something that has already become Skynet? Also, what if becoming Skynet is the least expensive option in the problem domain? That's only the equivalent of 22 million water crossings.
You just put in a function that checks if your algorithm is going to become Skynet in the next generation, before you actually apply the fitness function.
That's what I do.
And you don't want the evolution of Skynet to be impossible, just costly.
John -
Whatever you had linked to has moved on - there's nothing visible about Skynet there now - at least not from the USA!
Ian, I still get the xkcd page (with rollover - which reminds you that you want to minimize, not maximize, the function).
Ok, what's skynet? Is that the thing from the Terminator or some kind of ISP?
Nevermind. Spent 10 seconds to look it up myself. Does this mean I lose nerd-cred, both for not automatically knowing this (sorry, watched the first movie, liked it, thought the second was a lot of lame, and never watched anything else dealing with the Terminator), and for not googling it?
Heh at least it was in python.