Seite - 95 - in Limina - Grazer theologische Perspektiven, Band 3:2

95 | www.limina-graz.eu Sara Lumbreras and Lluis Oviedo | Belief networks as complex systems of ongoing experiences to guide future actions. All these functions are sup- ported by beliefs and the process of credition. The recent successes of AI have injected renewed strength into the com- puter metaphor. While the old computer metaphor, which attempted to assimilate the mind to a particularly efficient calculator, has proven un- successful, alternative approaches to this simile are starting to emerge, based on connectionist models rather than serial ones and flexibility and adaptability as opposed to predictability and design. Many are the thinkers who see in the recent achievements of AI the basis to substantiate claims for the possibility of creating even artificial consciousness (Kurzweil 2012). The remainder of this section explores the two most prominent AI tech- niques in this context: Artificial Neural Networks (ANNs) and Reinforce- ment Learning (RL). Since many newly published papers in the field of AI studies deal with ‘be- liefs’, it appears that such an approach might become fruitful when we try to untangle the intricacies of the believing process. Just as an example, browsing for the term ‘belief’ in the titles of research articles in the Journal Artificial Intelligence, we get 739 entries, i. e. published titles that include the word ‘belief’. Among published titles in journals or edited books in the field of AI, we find for example: Belief and truth in hypothesised behav- iours; Group Decision Making via Probabilistic Belief Merging; Lifted first- order belief propagation; Probabilistic Belief Embedding for Knowledge Base Completion; Causal Basis for Probabilistic Belief Change: Distance vs. Closeness; Probabilistic Belief Revision via Similarity of Worlds Modulo Evidence; Belief Systems and Partial Spaces. The brain as a pattern-recognition system Pattern recognition is arguably one of the main functions of the brain, and a particular field where the success of AI has been incontestable. Artificial Neural Networks are computer systems vaguely inspired by the neural pro- cesses in animal brains. We recommend the introductory work by Hassoun (1995) for a complete working guide, but will proceed to provide a stylized description of this tool for the purposes of supporting our framework. Networks are composed of several units (artificial neurons) that work to- gether to perform tasks defined by the programmer, such as classification Artificial Neural Networks and Reinforcement Learning