One way to address the growing gap between AGI capabilities research and safety research could be through adapting Street Epistemology (SE)—a Socratic dialogue technique—for AI safety outreach. As AI capabilities advance rapidly, traditional methods like lectures or debates often fail to meaningfully engage capabilities researchers on safety concerns, sometimes even reinforcing their skepticism. This approach would focus on creating reflective conversations rather than confrontational debates, potentially making belief updates more likely.

How This Could Work

The idea would involve partnering with existing SE YouTube channels, where practitioners film dialogues with strangers about their beliefs. AGI safety researchers could appear as guests, demonstrating how to discuss safety concerns through non-confrontational questioning. This could serve multiple purposes:

• Showcasing effective dialogue techniques for AGI safety conversations
• Attracting SE practitioners to apply their skills to AI safety
• Indirectly exposing capabilities researchers to safety arguments in a low-pressure format

The Socratic nature of SE might help bypass defensive reactions that traditional outreach often triggers. Instead of directly arguing against capabilities research, the dialogues would encourage participants to examine the foundations of their own beliefs about AI development.

Potential Benefits and Implementation

This approach could benefit multiple groups:

• Safety researchers would gain a new outreach method with lower risk of backfiring
• SE practitioners could expand their work into high-impact areas
• Capabilities researchers might become more open to safety considerations through this indirect approach

A pilot phase could start with safety researchers who have strong communication skills appearing on established SE channels. If successful, this could expand into training materials for SE practitioners and potentially lead to one-on-one dialogues between SE practitioners and AI researchers. The YouTube format offers scalability, as successful dialogues could reach broad audiences with minimal additional cost.

Comparison With Existing Approaches

Unlike technical papers or formal workshops, this method would be conversational and adaptable to individual perspectives. While existing outreach often preaches to the choir or comes across as confrontational, SE dialogues could engage skeptics where they're already active—such as on YouTube or other platforms. The interactive nature might prove more effective than one-way communication like blog posts, as it allows real-time adaptation to the participant's specific concerns.

This approach wouldn't replace other outreach methods but could complement them by reaching researchers who are resistant to more formal or direct engagement with safety arguments. The key advantage would be its ability to potentially shift perspectives without triggering the defensive reactions that often hinder productive discussion about AI risks.