AI Audio Cloning: Einstein Chatbot Gets a Voice

The Emergence of AI-Powered Deepfakes: A Digital Einstein

A novel application of synthesized media has surfaced: a digital representation of Albert Einstein, complete with an artificially recreated voice. This was achieved through the utilization of AI voice cloning technology, leveraging existing audio recordings of the renowned scientist.

Development and Collaboration

The startup, Aflorithmic, is responsible for the audio deepfake, having previously received seed funding earlier this year. The 3D character rendering is the work of UneeQ, another synthesized media firm, and the interactive chatbot is hosted on their platform.

Potential Applications and Concerns

Aflorithmic envisions this “digital Einstein” as a demonstration of the future possibilities within conversational social commerce. This suggests that deepfakes portraying historical figures may soon be employed for commercial purposes, a scenario industry experts have previously anticipated.

The company also highlights the educational benefits of bringing historical personalities to interactive “life,” albeit a virtual one. However, this “life” is an artificial approximation, and the voice isn’t a perfect clone.

The Role of Human Input

Digital Einstein’s voice creation involved collaboration with an actor for voice modeling. This was necessary to enable the chatbot to articulate concepts the real Einstein never would have considered, such as “blockchain.” Therefore, the project incorporates more than just artificial intelligence.

Matt Lehmann, Aflorithmic’s COO, stated that this project is a milestone in showcasing the technology for conversational social commerce. He acknowledged ongoing technical challenges and imperfections but expressed optimism about the direction of development.

Technical Improvements and Remaining Challenges

Aflorithmic has made significant progress in reducing the response time of the chatbot. Initially, it took 12 seconds to process input and generate a voiced response; this has now been reduced to under three seconds, approaching “near-real-time” performance.

Despite this improvement, a noticeable lag remains, preventing the bot from feeling entirely seamless.

Legal and Ethical Considerations

The creation of “digital clones” raises legal and ethical questions regarding data and image rights, particularly concerning living individuals. Consent and compensation are likely required before replicating someone’s likeness.

Historical figures, lacking the ability to object, present a different scenario. However, licensing rights may still apply, as is the case with Einstein.

Licensing and Artistic License

“His rights are held by the Hebrew University of Jerusalem, who are partners in this project,” Lehmann explained. He further clarified that the Einstein voice wasn’t a direct clone but was inspired by original recordings and films.

The voice actor involved was a devoted admirer of Einstein, and his performance successfully captured the character’s essence, according to the developers.

The Impact of Deepfakes

The impact of deepfake content is arguably more crucial than the sophistication of the technology itself. Context plays a vital role in shaping public perception, and even imperfect fakes can influence narratives.

Growing concerns about the potential for deepfakes to spread disinformation are driving regulatory efforts, particularly in Europe, where both companies behind “Digital Einstein” are based.

EU Regulations on Artificial Intelligence

A leaked draft of proposed EU legislation on “high-risk” AI applications includes specific provisions targeting deepfakes. Lawmakers are considering “harmonised transparency rules” for AI systems interacting with humans and those generating or manipulating media.

This could require future chatbots, like Digital Einstein, to explicitly identify themselves as artificial to prevent deception. This would avoid the need for users to discern authenticity through methods akin to a “virtual Voight-Kampff test.”

Current Status and Future Outlook

Currently, the Digital Einstein chatbot’s lag and clear labeling by its creators serve as indicators of its artificial nature. Aflorithmic aims to showcase its vision of AI-driven social commerce to potential business clients.

Further information regarding the EU’s proposed AI regulations can be found here: https://techcrunch.com/2021/04/14/eu-plan-for-risk-based-ai-rules-to-set-fines-as-high-as-4-of-global-turnover-per-leaked-draft/amp/