With a busy enough global pandemic for the past few months, you may have forgotten about this other apocalyptic danger, namely unchecked artificial intelligence. Unfortunately, an AI program commissioned on Microsoft's MSN.com has brought it back into the spotlight. The story centers on an article posted on MSN.com about Little Mix star Jade Thirlwall misusing a photo of another mixed-race group member, Leigh-Anne Pinnock. That would be a pretty serious mistake for a human editor, not only because of its horrible timing amid worldwide protests against racial injustice, but it takes on a whole new flavor when the automated processes behind this article appear. MSN.com does not do its own reporting, preferring to repurpose articles across the web and distribute the resulting ad revenue with the original publisher whose article was able to reach a larger audience. It's not surprising, then, that Microsoft thought much of this process could be automated, not just laying off hundreds of employees during a pandemic, but also implementing a publisher. AI that automatically posted news to MSN.com without human supervision.
not so smart after all
Racial bias in AI is a well-documented problem, with the huge data sets used to train artificial intelligence software often seeing these programs reproducing human biases, which, under the guise of algorithmic objectivity, race-focused organizations AI can easily miss. (For a truly dystopian example, check out this psychotic AI fashioned from Reddit data sets.) This bug shows what can happen when callous AI is entrusted to editorial oversight, requiring empathy beyond an algorithmic sense of reader interest. Most perplexing in this story, though, is the breadth of the role Microsoft felt comfortable assigning to an AI, with the ability to auto-publish stories without human curation, meaning no one had a chance to prevent this bug before it will arrive at MSN.com Readership. We also learned that if the article was quickly removed by human staff, a term we can't believe we should use, AI still had the option to cancel them and republish the same story. In fact, that's essentially what happened: with the press articles about the AI auto-exchange regarding their own mistakes, deeming them relevant to MSN.com readers. When contacted, a Microsoft representative said, "As soon as we became aware of this issue, we immediately took action to resolve it and replaced the incorrect image." However, we have yet to hear anything about the change in these AI processes. While AI software may seem like a practical replacement for newsroom roles, we've seen that AI applications aren't ready to take editorial responsibility from human hands and try to rush into the future. AI-powered will only hurt Microsoft's reputation and online reporting. Via The Guardian