Exploratorium: Deep Frank (AI)

Whisperer, Tinkerer, Designer

For years, I’ve been captivated by generative art and the computational processes that drive design and narrative development. As AI workflows and tools continue to evolve, they provide unprecedented opportunities to experiment with new modes of interaction, innovative research methodologies, and novel ways of framing our questions. The inspiration for Deep Frank came from a conversation at the Exploratorium about the institution's role in catalyzing action in the world.

I believe that knowledge alone is insufficient. My goal is to put knowledge into action, test it, and reflect upon it. Sisagreement as inevitable, I see success not in achieving consensus but in fostering functional and empathetic disagreement. Deep Frank is a series of AI experiments modeled after Frank Oppenheimer's approach to learning, compassion, innovation, and inquiry. This avatar serves as a reminder to remain curious and to embrace the spirit of inquiry.

 

DEEP Frank V2

May 2024

Deep Frank emulates Frank Oppenheimer's knowledge and curiosity about scientific and artistic phenomena. Be curious. Ask questions and question answers.

My second iteration of a Frank Oppenheimer AI experience. This time I built a GPT, using OpenAI, that has ingested all of the of the content I could find about Frank: papers he wrote, interviews, quotes, etc. The GPT has been given instructions on prompts for questions the user might consider, emphasizing the informal science learning approach that has taken shape at the Exploratorium.

You’ll need a paid version of ChatGPT to use this version of Deep Frank. My next step will be making it publicly available as a chatbot embedded on a site.

 

Deep Frank V1

April 2023

I was researching Frank Oppenheimer’s motivations to start the Exploratorium and found myself wondering -how he would answer the question, “Why do you think art and science are important today?” Frank passed away in 1985 so, as a substitute, I create this one-hour prototype of him giving an AI-generated, speculative, response to that question.

Ultimately I’d like to build this as an on-the-floor and online experience in which visitors can speak directly with Frank.

The spoken cadence is a bit stiff and the lack of mobility in his body is distracting but again, I made this in an hour.

  1. ChatGPT was used to create the speech.

  2. An audio sample of Frank speaking was uploaded to ElevenLabs to synthesize his speech.

  3. Connected an archived photo of Frank to the synthesized speech in D-ID to create the deep fake.