Although it may be difficult to determine whether someone on TV or in real life is telling the truth, as artificial intelligence and speech synthesis improves, you may not only have to attempt to determine a person’s motivations, but decide whether what you’re speaking to is human at all.
That’s the idea behind DT R&D’s prototype for a wearable AI speech detector called “Anti AI AI.” It employs Google’s Tensorflow machine learning software to analyze voice patterns and then let a user know whether he or she is actually listening to a human.
Though currently still a proof of concept, the idea would be to implement this technology in a Bluetooth-enabled cyberpunk earpiece, giving its wearer feedback via a thermoelectric Peltier plate which cools an area near the back of the neck when talking to an AI.
To accomplish this, DT R&D developed a custom circuit board based on a Simblee BLE module, along with an iOS app that streams audio to the Tensorflow neural network that has been trained on a database of synthetic voices.
A demo of the dystopian device shows Anti AI AI distinguishing between a real recording of President Donald Trump’s voice and a digital clone. You can read more about this project in the Australian creative agency’s blog post here.