Definition
The Child Machine is a concept introduced by Alan Turing in 1950. Instead of trying to program an adult-level intelligence directly, Turing proposed creating a simpler machine that simulates a child’s mind and then subjecting it to an appropriate “course of education” to develop an adult-level intelligence.
Why It Matters
It suggests that the path to artificial general intelligence is not through hard-coding knowledge, but through building systems that can learn and adapt from their environment.
Core Concepts
- Learning over Programming: The core of the Child Machine is its ability to learn from its environment and teachers, rather than having all its knowledge pre-encoded.
- Bootstrapping Content: The machine starts with a “blank slate” (tabula rasa) or minimal innate structures and builds its understanding of the world through experience.
- Trial and Error: Turing envisaged an iterative process of experimentation, teaching, and evaluation to find the “best” child machine.
- Expedited Evolution: While natural evolution produced intelligence through random mutation, the Child Machine can be improved by an intelligent “experimenter” who can trace causes for weaknesses and implement targeted “mutations.”
- Fixed Architecture: Unlike a Seed AI, which modifies its own architecture, a Child Machine (in Turing’s original sense) primarily develops its content within a relatively stable learning framework.