Amanda Overcash, a single mother from Texas, has mastered the art of balancing her full-time job in real estate with parenting and training AI models. At night, after her daughter goes to bed, Overcash transforms into an AI trainer, reviewing chatbot responses, transcribing audio clips, and labeling images from her kitchen table.
The Flexibility and Demands of AI Training
Overcash is part of a global, largely invisible workforce that underpins the AI boom. While some contract workers have had negative experiences, Overcash describes her experience as largely positive. The pay can be lucrative, up to $40 an hour, but it’s not without its challenges. Last summer, she earned nearly $8,000 in under three weeks by working long days and nights on a chatbot evaluation project. However, she warns that it’s not a ‘get rich quick’ scheme; the work is demanding, and juggling it with a full-time job can risk burnout.
Juggling Multiple Projects and Platforms
Overcash has spent over six years in the AI data industry, taking on various projects like ad moderation, transcription, and prompt evaluation. She works across multiple platforms, including Appen, OneForma, Prolific, Outlier (owned by Scale AI), and Amazon Mechanical Turk. The pay rates vary depending on the project and its difficulty. For instance, large language model projects usually pay around $20 an hour, while social media or transcription tasks can pay between $9 to $11 an hour.
Types of AI Training Tasks
Currently, Overcash is working on two main projects: transcribing casual voice memos and reviewing social media ads. The voice memos are often recorded in noisy environments like cafés or cars, making the task challenging. When reviewing social media ads, she watches videos or reads captions and answers yes/no questions about nudity, profanity, and misleading claims, among other factors. She finds this task to be one of her favorites because it allows her to work quickly without second-guessing her answers.
Challenges and Rewards
While the work can be rewarding, it’s not without its challenges. Overcash has worked on intense projects that require up to 16 hours of work per day. She recalls a chatbot evaluation project where she had to review chatbot answers to medical questions, political statements, and personal advice, flagging any misleading or unsafe responses. The strict time limits and audit processes can be stressful, and a single failed audit can cost freelancers access to work for the day or even remove them from a project entirely.
Balancing Work and Personal Life
Overcash has learned to set boundaries to avoid burnout. She doesn’t work on weekends, reserving that time for her daughter. Her hours are flexible; some days she works two hours, while others she works eight. Despite the challenges, Overcash finds value in her work. It has made her sharper and better at spotting issues or bias in language. The flexibility, steady income, and control over her time are what keep her going.
As Overcash puts it, ‘It’s not a fast way to make money, but if you get into a rhythm, it helps. It’s helped me pay bills, stay afloat, and show up for my daughter.’ Her story highlights the realities of working in the AI training industry – the flexibility, the challenges, and the rewards.
