Joshua S. Rule and Maximilian Riesenhuber are the scientists who have published their study describing a newly created AI software which will improve object recognition.
Humans are intellectually gifted species who have the ability to process information quickly. For example, if you give a pen and paper to a toddler he/she will end up drawing lines or dots on a piece of paper. Similarly, AI powered devices are expected to accept or reject objects.
Joshua and Riesenhuber have identified a particular limitation with computers as far as vision is concerned. Their analysis takes into account the fact that Computers are visually impaired and so, time and again they are to be shown same behavioral patterns to make an exact clone of the same.
The scientists share, “The various Generic feature sets were chosen as controls against which to compare the conceptual features. Based on prior work using GoogLeNet, these layers likely correspond to high-level visual cortex.”
This is a move from using low-levels of data to higher-levels, in order to keep computers at par with the human race. Although, the truth remains that human beings will never be replaced.