A lot is discussed about world models currently and the next step towards AGI. The idea for them are not new and are based on „The Nature of Explanation“ by Kenneth Craik. I highly recommend the book.
In this first post about this book I am going to give a high abstract view on his ideas. The text here is based on the book ISBN 978-0-521-09445-0.
At first the philosophical view on the world by great philosophers like Huem, Descartes, Berkeley are in his opionion way too – I would call it – selfish. He proposes for modeling problems of knowledge and explanations five attitude theories [on page 7 in the book]:
- A priorism
- Scepticism
- Descriptive
- Relational
- Causal
A priorism
This attitude asserts certain facts and principles to be self-evident or certain, and deduce a great body of supposed knowledge therefrom.
Scepticism
This attitude denies the legitimacy of these first principles, and questions some or all of the foundations of the belief in an external world and causal interaction.
Descriptive
This attitude defines theories, which assert that explanation is „generalised description“ but never tells us anything about the cause of events.
Relational
This attitude declare theories (represented by modern physics), which declare themselves to be uninterested in whether causal action between supposed ultimate units may be taking place, on the ground that things are unobservable and hence unpredictable; and that the aim of science is to find relations hence permit successful predictions to be made. The foundation of this method is the association of definite probabilities, smaller than unity, with events.
Causal
This attitude declares theories, which old that the events we see are the consequences of the interaction of external objects according to definite and certain rules.