The name "LExAu"
I will follow the example set by D.E. Knuth who makes the effort to explain the name of his projects (for example: TeX). Especially the pronounciation.
The name of the project, "LExAu", stands for Learning Expectations Autonomously. It is pronounced in a French fashion: \lek'sO\ (the link points to a page that lists the pronounciation symbols that are used by Merriam-Webster). The three
words that form the acronym represent the core principles that the project is based on:
- Learning
- There are a lot of ways to approach A.I. To name a few: Neural Networks, Inductive Logic Programming, Evolutionary Programming, and Computer Vision. The approach of the LExAu project is Machine Learning. The main feature of learning is that a learner accumulates knowledge about its environment. This is the first way to prove the statement in the first paragraph on this page wrong. You don't put stuff into a learning system, on the contrary: a learning system gets knowledge out of its environment.
- Expectations
- This keyword is meant to contrast with predictions. Predictions are black-and-white, they are also most of the time wrong. Expectations are more useful than predictions. You can bet on them, literally! This could turn out very helpful if we view learning as a betting game between a learning system and its environment.
- Autonomously
- This is the new element that this project introduces to machine learning: rather than a learning system that can learn a task, we strive for a learning system that is curious about its environment. An autonomous learning system may be constrained by internal limitations, but it is not dependent on its environment for teaching it. The environment does not supply losses or gains. This introduces a surprise factor that also aids in disproving the statement above.
Of course the autonomy principle implies that we have to study interactive systems. Curiosity implies actions and expectations imply responses, so let us create an A.I. that interacts…