The "model" presented for ML is low complexity, hence lives on the high bias, low variance side of the space of possible models.