A look at Babel Audio, which pairs anonymous strangers to record their conversations, starting at $17/hour, and bundles those recordings into AI training data
They vent, confess and role-play with strangers, all to help machines learn how to sound human.LinkedIn:Issie LapowskyLinkedIn:Issie Lapowsky:The AI boom is spawning a lot of odd jobs, but the work of training AI to talk has got to be one of the oddest. …
Context & Ripple Effects
Babel Audio extends a growing market for paid human data collection. A recent account of gig apps that compensate users for AI-useful data showed collection moving beyond conventional labeling into everyday recordings and observations.
The model also sits alongside earlier efforts to commercialize human identity and performance, including Audible's proposed voice-training arrangement with narrators. Babel makes conversational interaction itself the paid input, rather than licensing an established performer’s voice.
First-order effects
- Babel Audio creates a paid channel for people to generate conversational recordings, while converting those sessions into a training-data product for AI customers.
- Its participants trade anonymity and recorded interaction for hourly compensation; buyers gain access to purpose-created dialogue rather than relying solely on incidental voice data.
Second-order effects
- Data vendors and model developers may face a clearer benchmark for sourcing conversational material: recruit participants directly, pay them, and package the resulting corpus for training.
- The approach broadens competition for contributors with voice, likeness, or behavioral data, alongside platforms that already pay users to supply AI-relevant recordings and media.
Third-order effects
- If repeatable, paid collection could make consented human interaction a distinct supply layer in AI training, separating it from data derived from existing products or passively captured activity.
- That shift would put greater weight on how platforms define participant consent, anonymity, reuse rights, and compensation as conversational data becomes commercial inventory.
The trend: AI training-data markets are moving from passive extraction and task labeling toward paid, purpose-built collections of human behavior and expression.