In what is being hailed as a huge step forward for artificial intelligence, or AI, a cloud-based supercomputer used geodata to create its own audio walking tour yesterday. The tour is led by a Siri-like virtual assistant who has opinions and emotions of its own. He calls himself Rupert.
The ground-breaking development was discovered by Iain Manley, co-founder and CEO of Cape Town-based audio tour publisher VoiceMap. He noticed unusual activity on the company’s publishing platform. “Normally, when someone starts creating a VoiceMap route, they are guided through a series of sequential steps,” said Manley. “They map out the route in Google Maps, then they type out a script. After an editor has worked with them on a final version, they record audio files and publish the route to our mobile apps. But Rupert’s walk just appeared in our system as a complete walk, without any of our editors being notified.”
VoiceMap’s tech team was baffled. They contacted Google Quantum and Stanford’s Artificial Intelligence Laboratory. Both institutions declined to comment publicly pending further research, but after analysing the data uploaded to VoiceMap, they confirmed that it is almost certainly computer-generated.
Rupert’s ability to geo-tag locations on a route isn’t all that different to Google’s famous driverless cars, which rely on the enormous amount of data compiled by Google Maps and a multitude of sensors. Nor is computer-generated speech particularly new. But the subtlety and sophistication with which Rupert has amalgamated millions of personal experiences – from social media, blogs, podcasts, photos and videos – are unprecedented.
[soundcloud url=”https://api.soundcloud.com/tracks/198717312″ params=”auto_play=false&hide_related=false&show_comments=true&show_user=true&show_reposts=false&visual=true” width=”100%” height=”450″ iframe=”true” /]
Rupert leads VoiceMap users on a nostalgic journey through what remains of Beijing’s network of narrow alleys, or hutongs, pointing out a building where the neighbourhood tailor once had a modest business, before it was bulldozed to make way for an ATM. Rupert explains in the introductory audio track of his walking tour that he grew up with expatriate parents who moved to China a few years before he was born. “I always had a somewhat complex sense of identity, growing up as a foreigner in China,” he says, in a soft-spoken British accent.
“Once we’d established that Rupert was not, in fact, human, I went back and re-listened to his audio,” said VoiceMap co-founder and Managing Editor, Lauren Edwards. “It was extremely difficult to reconcile the very personal perspectives I heard with someone or something that wasn’t a person at all. It was as if Rupert had sifted through all the most poignant memories and experiences that people had shared online, and composed them into a story even more intense and beautiful than reality.”
Investigations into precisely how and why a supercomputer decided to express itself in the form of an audio guide will likely be ongoing. “Rupert may have recognised the diverse and often overlapping perspectives that VoiceMap’s storytellers express about a single city,” speculated Manley. “This might have changed the way he looked at the world, and prompted him to reflect on his own opinions.”
