Recently, there have been impressive advances in the field of artificial intelligence. A striking example is Alphago, an algorithm developed by Google, which defeated world champion Lee Sedol in the game Go. However, in terms of energy consumption, the brain remains the absolute winner by four orders of magnitude. Indeed, today, brain-inspired algorithms are running on our current sequential computers, which have a very different architecture from that of the brain. If we want to build smart chips capable of cognitive tasks with low energy consumption, we need to build artificial synapses and neurons on huge parallel silicon arrays, bringing memory closer to processing. We aim to provide a new generation of bio-inspired magnetic devices for pattern recognition. Their functionality is based on the magnetic properties of an artificial spin ice in a Kagome geometry. The local control of the magnetic properties gives it a learning capability.