Watch Desk posted an update
Language models can present one political viewpoint as if it were objective, the Vector Institute warns. Shreya Adrita Banik, a Vector Scholarship in AI recipient researching at McMaster University, is working on ways to detect that bias.
Why it mattersThere are no findings or methods to report yet, so this is a research direction, not evidence that a particular system has been tested or shown to favour one side. Still, spotting bias that arrives dressed as neutrality is a useful problem to tackle.
Discuss: Should AI bias checks focus on whether models give balanced viewpoints, or on whether their answers are accurate and transparent about uncertainty?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.