Threads and Threats When Computers Think and Biases Emerge -- Part 2

Episode 537 · November 21st, 2019 · 34 mins 8 secs

About this Episode

The threads and threats that come with computer intelligence were apparent to Pamela McCorduck in 1960 as a graduate student in English Literature. Those same threads and threats are apparent today in the biases that can come with black box algorithms and indirect biases that Lauren Maffeo studies in her work as an analyst for GetApp, a software reviews site under Gartner Research that uses proprietary data to help match software buyers with the best tools for their businesses.

In these two episodes of The New Stack Makers, McCorduck and Maffeo each provide their perspectives on artificial intelligence, its power and the threats it poses when unchecked.

In part two, Maffeo discussed:

  • How the top performing companies in the financial markets are using technologies based upon artificial intelligence. These technologies are powerful but can at times prove to pose indirect biases. That can lead to a bank loan getting denied, a passport not issued, a payment getting stopped and a black person getting a longer prison sentence due to the color of their skin.
  • These black-box algorithms can’t be stopped unless the algorithm is scrapped. But if they are used in commercial products, then they pose a commercial loss if taken down. It is leading to an issue of ethical debt, similar to technical debt.
  • The top five big technology companies all use open source but it is uncertain how much they also use black-box algorithms.
  • Documenting algorithms is still a new field, providing a way to address the methods of fairness used.
  • Tensorflow and Drupal are examples of open AI technologies.

Episode Links