Machine learning techniques are scoring remarkable successes in the weather forecasting community, owing to an immense wealth of weather observations, and question the relevance of traditional equations-based numerical weather models and data assimilation. Operational oceanography, particularly in the Arctic Ocean, may appear like a dispossessed little brother in comparison to weather forecasting. Still, machine learning techniques hold a significant potential to improve ocean forecasts as well. The presentation will give an overview of the activities at the Nansen Center that use machine learning to reduce the costs of the forecasts, correct their biases and will hopefully make them more user-friendly.