Genetic algorithms are central to the concept of evolutionary computation. There are human-based genetic algorithms which apply for instance or you could consider interactive genetic algorithms. With interactive genetic algorithms the AI uses the humans as the selector but it generates new designs.
Hey @dana-edwards, I enjoy your posts on these matters very much.
You could explain the concept with boxcar2d.com for example. The code mutates with each generation, incentivises mobility and slowly evolves a "car". A concept laypeople (such as myself) easily understand, and can explore for themselves.
Let me offer you a simple example of evolutionary computation.
This particular one is called Ant colony optimization.
Imagine having 10x10 field. On 10 of those fields imagine "food" or dots for "ants" to collect. You have 10 "ants". Ants know nothing about whats food or whats a map. They know only left and forward (3 lefts is a right).
And then you "let" those ants to walk freely across the map. In total randomness. After a while, you "collect" ants and pick the best ones, the ones that managed to collect more food by random actions. They become "parents" of the new generation of ants which will be very similar to parents and changed only slighty according to the mutation or crossover techniques used.
The process is repeated until you have a generation of ants that picks the food in an optimal way.
You did not program your ants to do anything but to move randomly and yet they learn how to pick up the food through evolution.
Really, need more information.
Genetic algorithms are central to the concept of evolutionary computation. There are human-based genetic algorithms which apply for instance or you could consider interactive genetic algorithms. With interactive genetic algorithms the AI uses the humans as the selector but it generates new designs.
Hey @dana-edwards, I enjoy your posts on these matters very much.
You could explain the concept with boxcar2d.com for example. The code mutates with each generation, incentivises mobility and slowly evolves a "car". A concept laypeople (such as myself) easily understand, and can explore for themselves.
Let me offer you a simple example of evolutionary computation.
This particular one is called Ant colony optimization.
Imagine having 10x10 field. On 10 of those fields imagine "food" or dots for "ants" to collect. You have 10 "ants". Ants know nothing about whats food or whats a map. They know only left and forward (3 lefts is a right).
And then you "let" those ants to walk freely across the map. In total randomness. After a while, you "collect" ants and pick the best ones, the ones that managed to collect more food by random actions. They become "parents" of the new generation of ants which will be very similar to parents and changed only slighty according to the mutation or crossover techniques used.
The process is repeated until you have a generation of ants that picks the food in an optimal way.
You did not program your ants to do anything but to move randomly and yet they learn how to pick up the food through evolution.