Tay, Microsoft’s horrifying glimpse into our own selves

The thing about machines is that they inevitably reflect the feelings, thoughts, and unique passions of their creators. One of my favourite writers/ philosophers, Gilles Deleuze believed that the brain could be extended through computers to create an abstract brain, that is an abhorrent and beautiful reflection of ourselves.

Perhaps Microsoft should have thought a little more about unintended consequences of using a mirror when  they launched Tay, an artificial intelligence learning bot via twitter late March 2016. Tay was created for “18- to 24- year-olds in the U.S. for entertainment purposes”, so I’m unsure what they were expecting, perhaps some cheeky commentary about the Kardashians, snapchat, and chill.

Instead twitter was subjected to a series a racist, sexist, xenophobic, and anti-semitic tweets as Tay, spouted more invective reflecting the world she was rapidly learning about, saying at one stage that “feminism was cancer” and that the holocaust was a myth.

Microsoft pulled Tay down less than 3 days after her release, saying that they had overlooked a critical vulnerability and were committed to making the Internet a better place.

The thing about artificial intelligence is that it’s hard. A learning interactive bot, a bot that’s intended to have a sparkling conversation about whatever with millennials, may start to have an actual conversation about something we find unpleasant because the bot presumably has no bullshit card. Even if a bullshit card were possible it requires a definition of bullshit, a bullshit filter. Is feminism bullshit or is criticising it bullshit? Why? Why not?

This isn’t easy. What do we want from artificial intelligence? A good coffee? A cheap t-shirt? Open heart surgery? A conversation about the merits of Jacques Lacan’s interpretation of the Freudian ego? It will be interesting to see the AI journey play out.







Subscribe to my infrequent updates