At the University Hospital of Giessen and Marburg 6,000 patients are waiting for a diagnosis of their rare conditions. Most patients have spent years bouncing from one doctor to another, building up huge dossiers of medical notes. Rare diseases typically take at least five years to correctly name, and sometimes up to 30, by which time it can be too late for effective treatment. "This is an inefficient, costly business," Dr Jurgen Schafer, who heads the German university's medical team, said at a media conference at IBM Zurich in October. "The computer is not going to replace the physician. But with this amount of data, it is completely clear that we don't need more physicians – we need more computer power."

Most patients have spent years bouncing from one doctor to another, building up huge dossiers of medical notes. Rare diseases typically take at least five years to correctly name, and sometimes up to 30, by which time it can be too late for effective treatment. "This is an inefficient, costly business," Dr Jurgen Schafer, who heads the German university's medical team, said at a media conference at IBM Zurich in October. "The computer is not going to replace the physician. But with this amount of data, it is completely clear that we don't need more physicians – we need more computer power."

The website kickoff.ai was launched on June 7 by three PhD students at the Lausanne Federal Institute of Technology (EPFL) and is targeted at football lovers. "All three of us are football fans and we wanted to apply machine learning to a new data set," Victor Kristof, one of the brains behind kickoff.ai, While several other football results prediction models exist, kickoff.ai The first one is modelling individual players' performances, which ensures more variables are incorporated when predicting an outcome of a match. Most traditional models only take into account the performance of the entire team and not that of its constituent team members separately.