Thanks to an impressive and chilling example that shows the point of artificial intelligence, fans of theoretical physicist Albert Einstein will now be able to talk to him. The technology, brought together by synthetic media, sound cloning and sound editing firm Aflorithmic and digital human experts UneeQ, makes it possible to chat with the famous scientist on the Internet.
The duo set out to celebrate the 100th anniversary of Albert Einstein’s Nobel Prize in Physics and to find a more interactive solution that would allow more information to be learned about the great physicist. He began work. The idea was to create something that could immediately answer any question about Einstein or his ideas.
Actually, this isn’t the first digital person Uneeq has created to answer questions. The AI, called Sophie, is also on the Digital Einstein website. The developers wanted to test their digital format with a name that was widely recognized, renowned for its intelligence, and people would be excited to ask their science and technology questions.
While the technology existed to create this reality-inspired avatar, a major problem remained: How do we know what Einstein’s voice is? According to Aflorithmic’s press release on Digital Einstein, “The answer is: most people don’t know and probably won’t care. What they care about is being able to easily understand (Digital Einstein) and feel like interacting with Einstein.”
Working with historical records, the team determined that Einstein had a heavy German accent and a high-pitched voice, saying his words slowly and as if speaking to a friend. Unable to work with low-quality recordings (the accent on the recordings was often too heavy to be heard on the Internet), they were unable to use their normal voice cloning methods.
Here, too, they drew inspiration from the field of cinema and instead crafted a character for Einstein with key features of his speech. The result was an Einstein with a German accent who spoke calmly in English.
The final challenge was to reduce the response time of Digital Einstein so that it could feel like he was speaking to a real person. This meant developing a system that could process whatever the speaker asked, in near real time, and answer it with an “Einstein voice.” After much effort, the team managed to reduce the reaction time from 12 seconds to less than 3 seconds.
If you want to ask questions or chat with Einstein, you can reach Digital Einstein here…
25366
