Could an AI App help us talk to the dead? | BBC News

Key Concepts

• AI in Film Production: Transformation of special effects, writing, casting, and image creation through artificial intelligence.
• AI Avatars/Digital Twins: AI-generated representations of individuals, designed for real-time interaction.
• Likeness Protection: Safeguarding an individual's name, image, and likeness from unauthorized AI use.
• Legacy Preservation/Digital Immortality: Using AI avatars to store personal information, thoughts, and stories for future generations.
• Grief Tech vs. Legacy Tech: Distinction between AI tools designed to aid in grief processing and those intended for long-term personal legacy.
• Consent and Agency: Ethical considerations regarding an individual's control over their digital avatar, especially after death.
• Video Generation Tools: AI software used to create cinematic content from text prompts.
• Prompting: The process of giving text commands to AI models to generate desired outputs.
• Bias in AI: Systematic errors or unfairness in AI outputs due to unrepresentative or skewed training data.
• Hybrid Filmmaking Model: An approach that integrates AI tools with traditional human-led creative processes.
• Storytelling Primacy: The argument that the quality and essence of a story remain paramount, regardless of the technology used to create it.
• Closed Data Set: A restricted collection of data used to train an AI model, ensuring it only accesses specified information.

AI's Transformative Impact on the Film Industry

The discussion opens by highlighting the rapid and pervasive transformation artificial intelligence is bringing to the film industry. This change extends beyond noticeable special effects to fundamental aspects like writing, casting, and the creation of lifelike images. AI is reshaping entire workflows and unlocking new creative possibilities. The program features Dr. Stephanie Hair, filmmaker Mark Chislac, and actor Caleb Worthy, co-founder of the AI company Two-Way, to explore these developments from various perspectives.

Two-Way App: Protecting Likeness and Preserving Legacy

Caleb Worthy, co-founder of Two-Way, explains his motivation for creating the app. During the SAG-AFTRA and WGA strikes, he became concerned about his likeness and work being used without permission, experiencing deepfakes and chatbots impersonating him. Two-Way was developed as a solution to empower individuals and protect their digital AI presence.

Key Features and Functionality:

• User Control: Only the individual can create an avatar of themselves. Users retain ownership of their name, image, likeness, and all associated data.
• Data Privacy: Two-Way explicitly states it does not train on or sell user data, aiming to "amplify your presence, not replace you."
• Avatar Creation Process: Users record themselves for three minutes. They then "train" the avatar's "brain" by uploading text documents, voice recordings, transcribing videos or podcasts, or even entire books. This creates a "closed data set," ensuring the avatar only accesses information the user explicitly provides and wants it to know, accurately representing them.
• Commercialization and Global Reach: The platform allows for commercialization. Caleb, with millions of social media followers, uses his avatar to interact with his global fanbase in 80 different languages in real-time, maintaining his authentic voice while communicating in their native tongue.

The company's recent ad, which garnered 41 million views, depicted a multi-generational interaction with an avatar, sparking significant debate.

Ethical Debates: Grief, Consent, and Misunderstanding

The Two-Way ad ignited a conversation about "talking to an avatar of the dead," raising profound ethical questions concerning consent, grief processing, and an individual's agency over their digital self after passing.

Caleb Worthy's Clarification:

• No Post-Mortem Avatar Creation: Caleb explicitly states that "you cannot create an avatar of anyone who has already passed on 2A because that person doesn't have agency over what their avatar says."
• Legacy, Not Grief Tech: He clarifies that Two-Way is "not a grief type platform." Its primary value is not for the generation grieving the deceased, but for future generations who would never have had the chance to meet that person.
• Personal Legacy Example: Caleb created his avatar, embedding his thoughts, perspectives, and life stories. This allows his great-great-great-grandchildren to interact with him and understand who he was, much like he wishes he could interact with his ancestors from 1825. The platform aims to preserve personal stories for generations.

Expert Perspectives:

• Dr. Stephanie Hair notes that the market's understanding of Two-Way is "slightly inaccurate." She emphasizes that it's about "legacy," allowing a living person to control and monetize their digital self for perpetuity. She finds this less problematic than "death tech" and sees value for historical projects, researchers, and family archives.
• Mark Chislac acknowledges the "Black Mirror" echoes in the ad and the public's concerns about prolonged grief. He draws a parallel to USC's decade-old project of recording Holocaust survivors' testimonies to create interactive holograms, highlighting the historical value of capturing life stories. He attributes public apprehension to popular culture narratives about AI and loss of agency.
• Mischaracterization: Caleb believes the ad's focus on legacy was misunderstood, as much of their prior content highlighted educational and monetization aspects for experts and celebrities. However, he stresses that everyone has a story worth telling, making the legacy feature universally valuable.
• Post-Mortem Control: In the event of the creator's death, control of the avatar defaults to the person in charge of the estate, unless specific instructions are left in a will.

AI in Film Production: Challenges and Limitations

Mark Chislac shares his experience testing various video generation tools for a mini-scene from an action movie, collaborating with a feature film director. While AI promises to build cinematic worlds from scratch with simple prompts, drastically reducing costs (from "tens, if not hundreds of thousands of pounds" to "just a laptop"), significant challenges remain.

Benjamin Baroot's Experience (Director of "Daddy's Head"):

• Lack of Control and Randomness: Baroot found the biggest challenge to be the lack of control and reasoning beyond prompting. He described the process as "so random," with "so much interpretation with prompting."
• High Failure Rate: To achieve 10 working shots, he had to generate 150 shots, indicating a highly labor-intensive and inefficient process.
• Bias in AI: The AI model "constantly wanted to make Mark white," highlighting issues of bias in AI training data sets and the model's attempts to continually change ethnicity.
• Impact on Creativity and Performance: Baroot noted the inability to tweak details like lighting and costumes, which are crucial in traditional filmmaking. He also expressed concern that AI "gets rid of you and makes you again," potentially destroying and remaking actor performances, questioning whether "brilliant directors are going to be wanting to work like this."

Mark explains that the inconsistency stems from how AI models interpret prompts, often leading to "happy accidents" rather than intentional design. While the technology is improving, its current randomness limits its widespread use in professional production.

Industry Perspectives: Jobs, Quality, and Bias

The discussion shifts to the broader implications of AI for the film industry, including job security, storytelling quality, and inherent biases in the technology.

Caleb Worthy on Industry Evolution and Job Protection:

• Historical Precedent: Caleb notes that past technological shifts (film to digital, rise of streaming) have historically widened access for storytellers. AI has the potential to continue this trend.
• Responsible Use: He emphasizes the need for responsible AI integration.
• Human Core of Storytelling: He argues that the "people who physically make the film and TV shows possible" – actors, crews, writers, casting, directors, producers, studio executives – are central to the industry. These roles "can't be swapped out or automated without potentially sacrificing the soul of storytelling."
• Quality and Livelihoods: Protecting these jobs is crucial not only for livelihoods but also because it "directly impacts the quality of what audiences see." Empowered professionals create better stories.
• Hope for Expansion: Caleb hopes AI will "expand opportunities and create more stories but not replace humans who are at the cart of the creative process."

Stephanie Hair and Mark Chislac on AI Bias:

• Root Cause: Stephanie identifies the "root problem" as insufficient diversity in AI training data. To solve this, more voices, faces, and representation of the "full spectrum of humanity" are needed.
• Privacy Concerns: This solution presents a challenge for individuals who prefer privacy or do not wish to participate in the AI world.
• Unequal Risks: Stephanie highlights that women, for instance, experience significantly more online harassment and are disproportionately victims of deepfake technology compared to men.
• Training Data Reflection: Mark attributes bias to the way models are trained and the companies behind that training, rather than solely the diverse user base.

Advice for Aspiring Filmmakers (Trev's Question):

Responding to a viewer's question about using AI for a short film promo, Caleb Worthy offers advice:

• Story is Paramount: "The key of storytelling is story." AI is merely a tool.
• Quality Over Technology: He uses the example of Toy Story, stating he would have loved it regardless of animation style (CG, hand-drawn, live-action) because "the story mattered."
• Collaborative Nature: Storytelling is a collective effort involving hundreds of people.
• Balancing Access and Protection: While AI can help storytellers without massive budgets, it's crucial to protect industry professionals.
• Focus on Narrative: For beginners, the primary focus should be on crafting a compelling story, as "AI is just a tool and stories are the things that will last forever."

Conclusion

The discussion concludes with a synthesis of AI's current state and future potential in creative industries. While AI offers unprecedented opportunities for expanding storytelling, reducing costs, and preserving personal legacies, its integration is a "work in progress." Significant challenges remain, including ethical dilemmas surrounding consent and grief, technical limitations like randomness and lack of control, and the pervasive issue of bias in training data. The consensus emphasizes a hybrid model where AI serves as a powerful tool to augment human creativity and expand access, rather than replacing the essential human element. The "human creativity and skill remain the gold standard," and the quality of the story, crafted by empowered professionals, will always be the most vital component. Responsible development, diverse data, and a focus on human-centric storytelling are crucial for navigating AI's transformative journey.