Ever since the 1950s, people have suggested that the easy way to achieve artificial intelligence is to build an artificial baby and have it learn from experience. Actual attempts to do this have always failed, and I think this is because they were based on the Lockean baby model.
This section concerns the design of a robot child that has some chance of learning from experience and education. We do not mean reprogramming, which is analogous to education by brain surgery. The instructor, if any, should have to know the subject matter and very no more about how the program or hardware works than parents know about the physiology of their children.
Consider designing a logical robot child, although using logic is not the only approach that might work. In a logical child, the innate information takes the form of axioms in some language of mathematical logic. 14
[McC79] and [New82] both discuss using logical sentences to represent the ``state of mind'' of a system that doesn't use sentences directly. We don't mean that here. We are discussing a system that uses logical sentences explicitly. If you don't like this approach, read on anyway and then decide how your favorite approach would handle the problems we propose to solve with logical axioms.
We will deal with just four innate structures among those mentioned in Section 2. These are the relation of appearance and reality, persistent objects, the spacial and temporal continuity of perception and the language of thought. They are all difficult, and we can't yet go beyond sketching the kinds of sentences that might be used by the robot child. The design of the child robot requires many more.