Zeynep Tufekci is fascinated by what she calls ‘machine intelligence’. She studies the different methods that algorithms are used to predict crimes, business trends and even elections. In her TED Talk Machine intelligence makes human morals more important, Tufecki discusses her work and the increasing importance of humans in the man-machine interface.
Zeynep tells the audience about her decision to study computer science. She was interested in math and science, and wanted to find a job quickly and not deal with many ethical dilemmas. Now, she says, computer scientists are building machines that make several ethical decisions, and the engineers and programmers might be building their own biases right into the software.
Metrics are easy, she says, when we’re working on the basic meat and potatoes engineering problems. We can tell if a bridge is structurally sound, if an automobile gains fuel efficiency, or if a mission to the moon successfully reaches its destination. Machine learning and artificial intelligence gains aren’t so black and white. Programmers give the programs lots of data but the data is unstructured, and more closely resembles a messy social media profile instead of a chart showing different bearing sizes.
Tufekci goes through several examples of machine learning during her talk – hiring profiles, clinical depression prediction, parole and recurring crime statistics, the computer Watson playing Jeopardy, and social media news feeds. Her basic conclusion is that the programmers who enter the massive amounts of data required to generate predictive learning machines have internal bias, and there’s not currently an easy way to remove that bias from the programs themselves.
As an engineer this talk fascinates me when it relates to product design and development. As we continually build smarter and more predictive features into our products, we need to ask if our own internal bias can be influencing the way that a component views the world. Tufekci now splits her time between Harvard, UNC, Princeton, and the New York Times.
(Images courtesy of TED.com and Pixabay.)