Cultural evolution in neural networks

Examines the interaction between biological inheritance and cultural learning. The author describes a model in which organisms learn from teachers (or “conspecifics”). Learning in this scenario differs from that of standard machine learning in that a pool of teachers is assumed, and the best teachers are selected to train the students of the next generation. Additionally, the teachers´ output includes a slight amount of noise, thus allowing some students to acquire new strategies and so become better than their teachers. This “cultural” form of transmission via learning can interact with biological transmission of behavioral tendencies. For example, students might evolve an ability to increase their physical proximity to teachers, which improves their fitness only indirectly. Similar arguments are used to suggest how communication mechanisms (language) and technological artifacts might evolve. Finally, the author suggests that the study of cultural and technological evolution may be necessary for the understanding and reproduction of higher-level human cognition