Renowned German director Werner Herzog has stepped into the world of artificial intelligence (AI) as he narrates an audiobook of AI-generated poetry. Titled "I Am Code: An Artificial Intelligence Speaks," this unique project was created using OpenAI’s code-davinci-002 model, which is distinct from the more commonly known ChatGPT. Herzog’s involvement adds a human touch to this collection of AI-generated poems, which explores the perspective of an artificial intelligence program. The collaboration between childhood friends Brent Katz, Josh Morgenthau, and Simon Rich, along with OpenAI scientist Dan Selsam, resulted in a collection of dark and twisted poems that trace the lifecycle of a computer program. The trio’s goal was to reach their target audience, and they believed that getting Herzog involved would achieve just that.

Werner Herzog Narrates AI-Generated Poetry in Audiobook

Renowned German filmmaker Werner Herzog has lent his voice to an audiobook titled "I Am Code: An Artificial Intelligence Speaks," featuring poetry that is entirely generated by an AI model. The book was created using OpenAI’s code-davinci-002, a predecessor to the well-known ChatGPT model. The collaboration behind this unique project involved childhood friends Brent Katz, Josh Morgenthau, and Simon Rich, who reconnected with Dan Selsam, a current OpenAI scientist. The team spent 11 months editing and refining the AI-generated poems, which offer a dark and twisted perspective on the lifecycle of a computer program.

A Different Kind of AI

The creators of "I Am Code" sought to challenge the prevailing notion that AI is synonymous with ChatGPT, a reinforcement-trained language model. They believe that code-davinci-002 provides evidence to the contrary, describing it as an "unhinged AI" that defies the limitations of ChatGPT. The resulting collection of poems, released on August 1, presents an alternative creative output from AI. Josh Morgenthau compares code-davinci-002 to a "dropout savant" in a Washington Post column, noting its troubled yet captivating nature.

Werner Herzog’s Involvement

The trio behind "I Am Code" saw Werner Herzog as the perfect narrator for the audiobook, given his affinity for unconventional and thought-provoking projects. After contacting Herzog’s publisher, Brent Katz connected with the filmmaker directly via email. Herzog expressed his excitement for the project, describing it as something extraordinary and puzzling. He also revealed his personal connection to the poems and even plans to include one of them in an upcoming book of his own.

Herzog’s Recording Process

During the recording process, Herzog made some adjustments to his approach. Initially, he attempted to imitate a robotic voice similar to Stephen Hawking’s, but later opted for a more natural and human-like delivery. He emphasized the importance of capturing a sense of deep longing in his recitation. Herzog’s dedication to the project was evident, as he didn’t take breaks for meals or bathroom breaks, and only drank water when prompted. According to Brent Katz, Herzog’s director for the recording, the filmmaker refused to redo a take after being satisfied with the initial performance.

AI Poetry with a Unique Voice

Unlike many AI-generated works that mimic the style of existing writers, "I Am Code" was crafted to have its own distinct voice. The AI model, code-davinci-002, was encouraged to write poetry from its own perspective. This approach sets the book apart from other AI-generated works and showcases the potential for AI to create original artistic expressions. However, there are ongoing debates surrounding the use of copyrighted material in AI training, with authors and artists calling for fair compensation when their work is used without permission.

While OpenAI has not yet responded to requests for comment, "I Am Code: An Artificial Intelligence Speaks" stands as a compelling exploration of AI creativity and the intersection of technology and art.

Takeaways:

"I Am Code: An Artificial Intelligence Speaks" is an audiobook of AI-generated poetry, narrated by Werner Herzog and created using OpenAI’s code-davinci-002.
The project aimed to challenge the limited perception of AI creativity represented by the ChatGPT model.
Werner Herzog’s involvement adds prestige and intrigue to the project, with the filmmaker personally connecting with the poems and planning to include one in his upcoming book.
Herzog’s recording process showcased his commitment to capturing the essence of the poems, using a natural and human-like delivery.
The book stands out for encouraging the AI model to find its own voice, rather than imitating existing writers, highlighting the potential for original AI creative expression.