Newsletter

Often the term Artificial Intelligence is being used for any algorithm that is cleverly thought of! Although the origin of these algorithms may be very intelligent, and the algorithms work in a smart way, we wouldn’t call them AI. Even algorithms that work within the field of AI, are rarely very intelligent.

The field of AI has more definitions. One definition is that the way the algorithm is working is taught during a training phase, which both saves a programmer time and probably ends up with a better result. This is also known as Machine Learning. When we state that we work with AI, this is what we mean.

Real, or “strong”, AI will make robots think and act like human beings. This is not what we do!

Wax on, wax off

To train an AI, you need to make training sets. A training set is a set of input values along with preferred outputs. It is up to the trained algorithm to figure out how to get from the input to the output. The better the algorithm is trained, the easier it will be to find the proper output to a given input.

The TAIL tool

The training sets contains a (preferrably large) number of labelled objects. The label states what type the object has and a trained AI will try to put the proper label on new objects. To help manually put labels on the objects in the training set, we have developed the Tool for Artificial Intelligence Labelling (TAIL). Although the proper name should probably be Tool for Machine Learning Labelling (TMLL), it is just easier to pronounce TAIL. The tool will display the objects (assuming they ar images), and let the user put a label on them. Labels could either be suggestions found by a previous used algorithm, og the labels set by other users, whatever helps.