Group items matching

in title, tags, annotations or url

While AI is a master of pattern recognition, algorithms can only operate on parts of the world that humans can precisely describe to it.

Hidden biases may be written inadvertently into the algorithms used to decide who gets a job interview or who qualifies for a loan or for parole. If a data set considers the word "programmer" closer to the word "man" than "woman," or if you build a system that learns from Wikipedia, where only 17 per cent of profiles of notable people are women, these biases will be perpetuated in the machine.

"How the Quest for the Ultimate Learning Machine Will Remake Our World"

Researchers at the MIT Media Lab, think that facial recognition software has problems recognizing black faces because its algorithms are usually written by white engineers who dominate the technology sector. These engineers build on pre-existing code libraries, typically written by other white engineers.

Machines, even virtual ones, have biases. They are designed, necessarily, to favour some kinds of data over others. Unfortunately, we rarely question the judgments of mathematical models and, in many cases, their biases can pervade and distort operational reality, creating unintended consequences that are hard to undo.

What happens when we teach a computer how to learn?