Software individuals are software systems that are able to model their own structure, to export software modules and fragments to other individuals and to import it from them, to reproduce in order to produce new individuals, and to 'educate' younger individuals or to 'learn' from older individuals. Furthermore, all of this occurs on a cognitive level; we are not talking about neural networks or genetic programming, but about knowledge-based systems with these capabilities.
We first developed these ideas in the late 1970's, as presented at the IJCAI 1979 conference (reference below). After some years of maturation, the same concepts were realized in an experimental implementation called the Software Individuals Architecture that was presented in 2001 and in 2003.
More recently, the SIA implementation has been superseded by the experimental Leonardo system which is presented under another menu item on the CAISOR website. Leonardo has picked up many of the concepts and designs from SIA but it also differs in many ways.
| 1979-001 | Erik Sandewall: Biological Software. Proceedings of International Joint Conference on Artificial Intelligence, 1979, pages 744-747. |
| 2001-001 | Erik Sandewall: On the Design of Software Individuals. Electronic Transactions on Artificial Intelligence, vol. 5 (2001), pages B:143-160. |
| 2003-001 | Erik Sandewall: A Software Architecture for AI Systems Based on Self-Modifying Software Individuals. Proceedings of International Conference on LISP, 2003. |