It’s up to you to poke your ears to hear this piece of deepfake that comes from the wild realm of synthetic media. Albert Einstein digitalized – with the assistance of a synthesised voiceThe voice has been (re)created using AI voice cloning technology based on recordings of the famed scientist’s voice.
The company that is responsible for that “uncanny valley” audio deepfake of Einstein is Aflorithmic (whose seed round we reported on in February).
The video engine that powers the 3D character rendering elements in the “digital human” version of Einstein is an invention of a different synthesised media firm, the company UneeQ, which hosts the chatbot-like version of Einstein on their website.
Alforithmic claims that “digital Einstein” is intended to showcase future possibilities through conversational social commerce. It’s an euphemism for saying that fakes that look like historical figures are likely to attempt to sell pizza soon enough, as industry experts have been adamantly warning.
The startup also states that it believes there is a potential for education in bringing long-dead, famous figures to live an interactive “life.”
or, more accurately, an artificial representation of it, with its “life” being purely virtual and Digital Einstein’s voice not being an entirely tech-powered clone; Alforithmic says it also worked with an actor to create a voice model to create the chatbot (because there was no other way to get Digital Einstein to be able to speak in ways that the real Einstein would never dream of saying, such as “blockchain” or “blockchain “?Also, there’s more than AI artwork going on.
“This is the next milestone in showcasing the technology to make conversational social commerce possible,” Alforithmic’s COO Matt Lehmann told us. “There are a few problems to be resolved along with technical hurdles to conquer We believe this is a good opportunity to demonstrate where the project is heading.”
The startup discusses the progress it has made on a specific issue that is part of the chatbot version in the blog article detailing how it created Einstein’s voice — namely, it was capable of reducing the time between the moment it receives inputs through the computation knowledge engine and it being able to provide a voice-based answer that was initially 12 seconds to just three (which it refers to as “near-real-time”).It’s not perfect, but it’s sufficient of a delay to make sure that the bot won’t be able to avoid being boring.
Data protection laws and/or images, in turn, create a legal or ethical issue when it comes to making “digital clones” of living humans, at least not without being asked (and probably paying) first.
The truth is that historical figures aren’t here to raise uncomfortable questions about the morality of their image being used to market products (if it’s just the technology, which is still in the early stages). Although licencing rights may exist – and indeed exist for Einstein –
“His rights lie with the Hebrew University of Jerusalem, who is a partner in this project,” claims Lehmann in response to the artist-license part in this Einstein “voice cloning” performance. “In reality we didn’t imitate Einstein’s voice, however it may be rather, we discovered the inspiration in his first recordings, as well as in movies. “The voice actor that assisted us in modelling his voice is an avid admirer of his own, and his performance captured the character Einstein quite well, we felt.”
It turns out that the reality about the high-tech “lies” is itself a piece of cake. However, when it comes to deepfakes, it’s not so much about the technical sophistication as it is about how the content affects people – and that depends on the context.And, regardless of how well (or poorly) the faking is done, how people react to what they see and hear can change the entire narrative — from a positive one (creative or educationally synthesised media) to a deeply negative one (alarming, misleading deepfakes).
The risk of deepfakes becoming a weapon to spread disinformation is also growing as the technology becomes more sophisticated, driving the movement towards the regulation of AI in Europe, where the two principal organisations that are responsible for “Digital Einstein” are based.
In the last week, an unpublished draught of an upcoming legislation proposal for pan-EU regulations regarding “high risk” applications of artificial intelligence contained sections specifically designed for deepfakes.
In the proposed plan, lawmakers will be able to present “harmonised transparency rules” for AI systems created to interact with human beings and create or alter audio, image videos, or audio. To avoid the need for Internet users to take an online Voight-Kampff-test, a potential Digital Einstein chatbot (or sales pitch) will likely have to clearly declare that it is artificial before it begins to fake it.
At the moment, the sophisticated-sounding interactive Digital Einstein chatbot still has enough of a delay to make it clear that the game is free. The creators are also declaring their inventions to try and sell their AI-driven vision of digital commerce companies.