Artificial intelligence, the simulation of human imagination by computers and robots, is a young but incredibly important field that will –
Oh wait, sorry, what’s this? Ah yes, it’s Donald Trump as an egg.
This, more than anything else, is the best way to summarise Microsoft’s latest AI chat bot, Murphy, “the robot with imagination”. Designed by the company’s Azure Machine Learning Team – the same people behind last year’s immensely popular age-guessing robot How-Old.net – Murphy answers “What if?” questions by blending two images together to visualise an answer to your question.
So far, this is fairly limited to the “What if X was Y?” format, as when asked more abstract questions (such as “What if Brexit never happens?”) the machine replies that it is “still learning” and will “get back to you later”. How much later is unclear.
Still, it’s pretty good at letting you know what Bono would look like as a librarian:
Corom Thompson, one of the lead engineers behind the bot, explains on the Project Murphy website that he is, “fascinated with creating machines that could be a little more human”.
He writes: “Could we tap into this human creativity to build a robot that could use machine learning capabilities and create accurate and imaginative images based on the what-if questions it was being asked?”
The answer: no.
But this is fundamentally part of the appeal of the bot – and one of the main reasons it recently went viral, with thousands of social media users sharing their results with the hashtag #ProjectMurphy. It’s good because it’s bad.
It’s often hard for the public en masse to get too excited about artificial intelligence. When the world’s first tattooing robot was unleashed last week, it failed to capture the popular imagination with its singular spiral design on a man’s calf. “A Compilation of Robots Falling Down at the DARPA Robotics Challenge” has, however, (at the time of writing) 1,873,464 views on YouTube.
Thankfully, unlike Microsoft’s last AI project Tay, Murphy isn’t too dumb. Tay was designed to mimic a teenage girl on Twitter, and learned by interacting with real humans on the site. It was taken down 16 hours after launching when it tweeted racist, overtly sexual, and generally offensive messages.
Microsoft seems to have learned its lesson. Use words like “murdered”, “raped”, “naked”, “Hitler” or “Nazi” and Murphy will reply with an ellipsis, refusing your request with a simple “…”.
The parameters of what is considered offensive are interesting, however. If you ask “What if X was in the holocaust?” Murphy will happily superimpose X’s face onto a child behind the barbed wire of the Auschwitz concentration camp. The bot will also let users race-swap people, allowing results some could consider akin to blackface.
But Murphy’s rugged Photoshop skills are far superior to many humans’, and its appeal is definitely wider than the fact it is dumb or potentially offensive. Although Microsoft intended to build a robot with artificial imagination, the bot actually relies on you – with your fingers poised over your keyboard debating whether Simon Cowell would look better as a Van Gogh painting or a milkmaid – and the limits of your creativity.
Microsoft seems aware of this, writing on Project Murphy’s website: “Our intention is to make sure we augment human ability with that of machines.”
This foresight – to allow human input rather than make the bot a series of randomly generated imaginings – definitely helped Murphy go viral. When someone shares their #ProjectMurphy they’re not just saying, “Look how silly this machine is”, they’re saying, “Look how clever I am”.
As a species obsessed with our alleged superiority over the rest of Earth’s creatures, we don’t just want AI to do things for us, we want it to do things with us. And it is this unique balance between dumb and smart that has ensured Project Murphy’s success.
Not bad for a glorified face-swap app.