Deepfake: deep learning + fake

Deepfakes use artificial intelligence (AI) tools to simulate people doing or saying things they never did or said, or appearing in fabricated situations without their consent.

Deep learning refers to machine learning techniques that use artificial neural networks to form relationships between input and output data. These neural networks consist of artificial “neurons” (very loosely inspired by neurons in the brain) organized into layers that are trained on a dataset to associate particular input data with particular output data. For example, they are trained to recognize patterns in the data that could help distinguish features of a person’s face or detect an anomaly on an x-ray image. Deep learning approaches utilize more layers than other artificial neural networks and constitute a set of practices aiming to achieve robust and complex results from these multi-layered networks.

Reference For more on deep learning, see John D. Kelleher, Deep Learning (Cambridge: MIT Press, 2019), 1-37. Background theme image, “Deepfake” by Lightspring, from Shutterstock.

3 of 25