This activity is not simple. So, when a local-ZM is deactivated, it has to store the conditions, to be able to resume when it takes control again. There are problems associated with this activity. Some of the information can be lost or the external reality may evolve in the mean time so that the stored information will be of no use. In this way, any model, which takes control of the being, has to initialize before being able to regain full control. This activity of initialization is very complex and in some situations it might contain errors. Thus, it is rather difficult to do many activities in time- sharing. There is also a basic problem associated with the term "knowledge". As we know, the knowledge is associated with the predictions of a structure of models. So, the knowledge is associated with the structure of models and not with the external reality, as we'd like it to be. We should never ever forget this thing. Even more, knowledge is a non-sense if we do not declare the structure of models. Example: in any positive science, it is usual to say that something is true based on a specified theory (model). HOW M-ZM MODELS ARE MADE For a given external reality, the brain makes a structure of models, using information taken from the external reality or from other models. We will see how this function works in a specified situation: how a new M-ZM is made in interaction with a new external reality. This function is described for a normal and mature brain. The term "normal brain" will be treated later. Here, a "normal brain" is a brain, which is able to work as it was already described in the section of hardware facilities. A mature brain is a brain, which has enough YM and ZM models made during a long time of interaction with the external reality. An image is an information which is received as it is, in the same way as it would be generated by a TV-camera for instance. This kind of information, without any meaning in fact, has to be integrated by the brain as an image- model. As we already know, M-models have to find some entities in that image. They start by making a 3D-image. This is possible in a rather easy way because almost all beings have two eyes. So there are two plane images and M-models will make a 3D-image. Now, the basic problem is that from a 3D-image it is not an easy task to identify the entities. M-models will use any supplementary information associated with this 3D-model, as color, contrast, brightness, the movement of some entities and so on. Anyways, M-models have to associate entities to YM-models. This process could be affected by mistakes, but, because M is a model, there will be a lot of crosschecks that will allow to discover and correct some of the mistakes. For instance, if something round is discovered, it could be an apple (YM- apple) or a ball (YM-ball) or anything else. Once a possible entity is associated with a YM, the M-model will predict how this YM interacts with the other YMs of the model.