Right from old-fashioned matchmaking to modern dating services, romantic matchmakers have focused on what singles themselves desired when they assisted them in the hunt for their soulmate. In other words, there has been nothing decisively new under the sun for several hundred years. But there is now. At the request of DR3, researchers at DTU Compute have developed a self-learning algorithm and sent it in search of the recipe for a good relationship.
Based on very large quantities of data about 1,334 people in well-functioning relationships, the algorithm analyses the prerequisites for a good relationship:
“The algorithm teaches itself what the ingredients are in a good relationship, and how they are to be mixed.”
Professor Jan Larsen, DTU Compute
“The algorithm receives a huge amount of information about each individual person in each of the 667 relationships, for example about food and transport habits, childhood town, height, number of brothers and sisters, pets, consumption patterns, and much more—including things that are not normally regarded as relevant to our choice of partner. The algorithm then looks for a pattern in the relationship based on the information. In this way, the algorithm itself learns what the ingredients in a stable relationship are, and how they are to be mixed,” explains the creator of the algorithm, Professor Jan Larsen from DTU Compute.
Alpha Zero of love
The experiment differs from previous attempts to identify what makes a good relationship in that the algorithm itself learns what it is to look for. All existing notions of romance and soulmates have thus been disregarded. And as the algorithm does not have a love life of its own, it can start right from scratch and draw its own conclusions based solely on data. Jan Larsen explains:
“Artificial Intelligence (AI) is solving more and more highly complex problems. For example, the Alpha Zero algorithm from the Google-owned company DeepMind has taught itself to play chess at a high level. We’ve used a simpler self-learning algorithm for learning ‘the game of love’.
In line with the algorithm becoming a leading expert in love, it uses its knowledge to help the participants in the TV programme find the perfect partner.
Source : DTU