A great piece in Nature about how to use AI in research – wisely:

It is crucial for researchers to fully understand the training and input data sets used in an AI-driven model. This includes any inherent biases — especially when the model’s outputs serve as the basis of actions such as disaster responses or preparation, investments or health-care decisions. Data sets that are poorly thought out or insufficiently described increase the risk of ‘garbage in, garbage out’ studies and the propagation of biases, rendering outcomes meaningless or, even worse, dangerous.

There’s lots more in this fantastic article. Read it here.