Artificial Thought How To
| |
We need a rather general algorithm in order to implement a thinking process. A general principle applied to artificial thought and intelligence would be better than direct attempts to solve specific problems.
Among available techniques, neural networks are well suited for such a purpose: starting from a very simple basis, their versatility is virtually infinite. How to focus on the main features? How to give up at least for a while the less essential aspects?
|
Artificial Thought components
| |
What exactly do we mean by thinking?
Take a certain amount of data from a present situation, add experience acquired from past situations, and compute a response. This response is supposed to bring effects in accordance with predetermined goals. This response and the observed actual effects will be associated as a new experience to be reused in the future. Thought itself shall be memorized as a part of experience. If the current state meets all predetermined goals, then thought power is available for a "free ride" through the many paths of linked recollections.
Otherwise, an urgent need for reaching a predetermined goal focuses thought flow on this task. Artificial thought components shall comply with this description.
|
|
Elemental components of intelligence
| |
Stop building artificial intelligence cathedrals from sand and water. Let us come to elemental components of Intelligence, that is to say elemental components of General Artificial Intelligence. Let us come to elemental bricks and cement. From these elemental bricks, we will build stronger and higher. What exactly is intelligence? Instead of starting a sterile debate, take all usual definitions as opportunities for crossing different approaches. Common sense is sufficient enough to provide plenty of examples of intelligence. From these very different examples, what are the common characteristics? What are the common ingredients?
- Perceive a given universe
- Associate and record
- Rate schemes
- Reuse successful schemes
- Highlight exceptions from previous records
- build concepts
- Act on the perceived universe.
|
Specify, Brainstrom, Mathematics
Specifications will take a main part in designing an artificial thinking machine. Then, bright ideas will be welcome, in order to meet these specifications. Then, mathematics will be helpful. A thinking process shall be as simple as... a boomerang.
Will mathematics give rise to artificial thinking? Not directly. No need to know mathematics to design a boomerang (see Wikipedia).
Such a so simple and so wonderful flying machine comes from people who didn't know anything to mathematics. However, if you try and enhance an already existing boomerang, mathematics and physics will be helpful in order to spare time. Boomerang seems to be rather far from Artificial Intelligence. Let us admit it. Even in good old-fashioned artificial intelligence field, mathematics are sometimes left aside: Fuzzy Logic comes from mathematics but fuzzy controllers don't rely on mathematical modelling to control processes. They rely on human expertise. Browse this quick interactive tutorial about Fuzzy Logic.
|
|
