Page - 97 - in Limina - Grazer theologische Perspektiven, Volume 3:2

97 | www.limina-graz.eu Sara Lumbreras and Lluis Oviedo | Belief networks as complex systems weights that lead to very few errors. This would mean, in our example, that when the ANN is shown new pictures, it can accurately distinguish between car and non-car elements. In general, when there are more hidden layers in an ANN, the network can carry out more complicated processes. Convolutional ANNs, and other types of deep networks (in the sense of having many layers) have proven to be especially efficient in performing complicated tasks such as image processing. Here, the layers can be understood as providing a hierarchical structure for the patterns that are recognized abstraction. This hierarchical structure has been linked to the functioning of the brain, which some au- thors (notably Ray Kurzweil) have defined as a pattern-recognition system (Kurzweil 2012). According to Kurzweil, the brain is structured in pattern- recognition units that are activated when exposed to a similar stimulus. For instance, they might recognize a small black line on a white paper. Then, several of these patterns combined might be recognized as a larger pattern. Following this example, they might respond to the letter “A”. Several of these letters might form the word “APPLE”, which triggers the concept of the fruit, and so on. We can understand a different problem, forecasting, as a form of pattern recognition. If we feed time series of rain inflows to an ANN, it can learn to predict how much rain will come next, because it recognizes the patterns in the data that reflect the season or any particularities of the region that is being analysed. The renewed brain-computer metaphor assimilates the brain to a pattern- recognition system. There is merit in this metaphor, which appears closer to the truth than the dated brain-computer simile. The animal brain is far from following a deterministic code. There are indeed some functions of the brain that could be well described as pattern recognition. Belief systems can also have a pattern recognition function and act as classifiers, a task that can be understood as establishing what something is and what it is not. In addition, belief systems can also be used for forecasting, as they provide a model for the world that help us anticipate our actions. There are functions of the brain that could be well described as pattern recognition.