Machine Learning: Our growing ability to un-pixelate images will do a world of harm

Think about all the things that are pixelated to protect our identity. License plate numbers are routinely blurred, as are the faces of children and victims. When journalistic images of crimes are deemed too disturbing, scenes of horror are edited to preserve the dignity of those involved. Now though, researchers from University of Texas at Austin and Cornell University are calling the practise out: as machine learning comes on in leaps and bounds, simply muddling a series of small squares is no longer enough. Today, our deep learning tools – which use computer algorithms to crack codes – have become so advanced that Texas and Cornell researchers identified a number of obscured faces and numbers.