last update: 2019 04 19
index
contact

Intelligence

Intelligence is movement of things.

Useful intelligence and useful work is the same.

Even the most primitive machine is intelligent.

Artificial intelligence

IMO artificial general intelligence (AGI) is possible with machine learning by current technologies like e.g. neural networks or evolutionary algorithms.

Neural Networks and Deep Learning
Are neural networks the future of AI? How important are they?
From Artificial Intelligence to Artificial Consciousness | Joscha Bach | TEDxBeaconStreet (2016-12-13).

EVOLUTION IS THE NEW DEEP LEARNING.
Neuroevolution—Improving Deep Learning with Evolutionary Computation.
Solving Hard Problems with Evolutionary Computation.

IMO these aspects are sufficient and maybe required for useful AGI:

Drive

The drive to think and to do.

There is no free will.

Prediction

The prediction of value or values W2 representing the world in time T2 based on value or values W1 representing the world in time T1.
Action values are parts of the values representing the world.

Prediction is implemented as a function that leads from world value W1 to world value W2.
The ability to predict equates understanding: Knowing what enforces what.

Fantasy

A fantasy is a prediction where the actual world value was modified by the addition of action values.
It is faster and safer and cheaper if the process of trial happens as computer simulation (fantasy) and not as trial in the real world.

Related: Aphantasia: Ex-Pixar chief Ed Catmull says 'my mind's eye is blind' (2019-04-08).

Judgement

Judgement of predictions based on how well they can predict the world at time T.
This judgement is based on the comparison of the predicted values and the actual values based on sensor data.

Judgement of fantasies based on the state of the world that should be achieved.
Fantasies that lead to the desired state of the world should be confirmed and executed.

Action

Actions to modify the real world and thus the values that represent the real world for the AI.

For the AI, an action is represented as action value among the values that represent the world.
A program executes the actions represented by action values.

The ability to predict comprises and equates understanding of action values.