ML requires training a model to recognize inputs. It’s impossible to create one ML Model to recognize all possible real world situations, so it’s necessary to continue collecting data, “retrain” the model, and redeploy the updated model.
For anyone who’d rather not ship all of their information to someone else’s cloud, we’re still years away from real ML capabilities for our own devices.
If anyone knows of an available ML solution that can perform continuous training/learning on a local device, I’d love to hear about it.