The Potential of Multimodal On-Device AI in Content Censorship for All Ages: A Case for Keeping Section 230

The evolution of artificial intelligence has brought forth innovations that extend far beyond traditional computing, touching on aspects of digital life that were once thought unmanageable. One such breakthrough is the development of multimodal on-device AI capable of content censorship for both children and adults. This innovation not only holds the potential to safeguard users from inappropriate content but also provides a robust argument for retaining Section 230 of the Communications Decency Act, which shields companies from liability for content posted by users.

Understanding Multimodal On-Device AI

Multimodal AI systems are those that can process and analyze multiple types of data simultaneously, such as text, images, and even videos. By integrating these capabilities on devices, AI can operate in real-time without relying on constant internet access. This ensures privacy, speed, and customization in content moderation.

An example of such an AI system is Google's Pixie. Originally designed for tasks like photo organization and enhancement, Pixie's capabilities can be extended to include content moderation, providing a powerful tool for filtering both text and images based on predefined content ratings.

Implementing Content Ratings

Using standardized content ratings akin to those used by Hollywood—G, PG, PG-13, PG-17, M, R, and NR—an on-device AI can categorize and filter content appropriately. Here is a proposed structure:

• G (General Audiences): Content suitable for all ages, containing no nudity, strong language, or intense themes.
• PG (Parental Guidance): May include mild language and mature themes, allowing depictions of bikinis and swim trunks.
• PG-13: Permits non-sexual nudity but restricts full-frontal nudity and explicit content.
• PG-17: Allows full-frontal nudity but excludes sexually explicit acts.
• M (Mature): Permits depictions of erections and touching but excludes sexual intercourse.
• R (Restricted): Allows explicit sexual content.
• NR (Not Rated): Automatically filters illegal and harmful content, such as child sexual abuse material (CSAM), hypnosis, and sexual propaganda.

Benefits of On-Device AI Content Moderation

1. Enhanced Safety for Children and Adults: By filtering out inappropriate content in real-time, users—especially children—are protected from exposure to harmful materials. This system also prevents them from inadvertently sharing explicit content, addressing issues like the production of CSAM among minors.
2. Maintaining Internet Freedom: Users have the option to disable these filters, ensuring that the internet remains a space for free expression. The choice to filter content is personal, preventing blanket censorship while offering protection where needed.
3. Reduction of Online Threats: By blocking scams, grooming attempts, erotic texts, nude requests, and death threats, the AI system contributes to lowering suicide rates and enhancing overall mental health.

Addressing Section 230 Concerns

Section 230 of the Communications Decency Act has been a pillar of internet freedom, allowing platforms to host user-generated content without being held liable for it. Critics argue that it enables the spread of harmful content, but multimodal on-device AI presents a solution that negates this concern. By ensuring harmful content is filtered out before it reaches the user, platforms can continue to operate without the fear of legal repercussions while providing a safer online environment.

Conclusion

The integration of multimodal on-device AI for content censorship represents a significant advancement in digital safety. Systems like Google's Pixie can be adapted to filter content based on standardized ratings, protecting users from harmful material while preserving the essence of free speech. This approach not only addresses concerns around Section 230 but also offers a more nuanced, effective solution than alternatives like requiring photo IDs for accessing adult content or banning minors from social media. By leveraging the power of AI, we can create a safer, more inclusive internet for everyone.