The rise of generative artificial intelligence has led to the proliferation of chatbots powered by large language models (LLMs). These chatbots, including Google’s Bard, OpenAI’s ChatGPT, and others, have seen rapid improvement in their performance, thanks in part to the work of data annotators.

The TikTok phenomenon

Jackie Mitchell, a college graduate and TikTok influencer, stumbled upon data annotation as a side hustle to supplement her income. She began posting “day in the life” videos on TikTok, sharing her experiences with data annotation without disclosing the specific site she worked for. Her videos attracted a significant following, inspiring others to explore this opportunity. This trend, referred to as the “Jackie Mitchell effect,” has brought an influx of young women into the world of data annotation.

The power behind the scenes

Data annotation involves tasks like rating and describing AI model inputs and outputs to aid in machine learning. It encompasses a range of projects, from editing and fact-checking chatbot responses to more complex tasks like computer coding and language translation. Data annotation is crucial in refining AI models, ensuring accuracy, and enhancing their performance.

Steady income for side hustlers

Contrary to the historically low wages associated with data annotation, Mitchell and others have found it to be a lucrative side hustle. Mitchell’s hourly rate has increased from $17.50 to $20-$30 over two years, while Brin, another contributor, has earned about $1,000 in a few months. The flexible nature of data annotation work appeals to young women looking for extra income, often with unpredictable schedules.

While the popularity of data annotation among young women is promising, it raises concerns about potential biases in AI models. Relying on a single demographic for training data could lead to skewed outputs, akin to the issues seen in medical AI models trained on data from one race. Data annotation also faces challenges related to quality assurance, as paying for annotation doesn’t guarantee high-quality work.

Balancing automation and human involvement:

Some advocate for automating data annotation entirely to reduce costs and speed up the process. However, this approach can introduce errors, reinforce mistakes, and hinder the detection of unethical behavior. Preventing annotators from using AI tools for tasks is challenging without effective monitoring.

Despite its popularity, data annotation may not be a long-term career path for many, including Mitchell. The unpredictable nature of side hustles and shifts in supply and demand could lead to changes in the field. As data annotation continues to evolve, it remains uncertain whether AI models can fully replicate the human experience in creativity, writing, or art.

The emergence of young women on TikTok as data annotators has brought a unique perspective to AI chatbots. While concerns about potential biases and the future of data annotation persist, the “girlies” have left an indelible mark on the AI landscape. As AI continues to evolve, it is clear that humans will always have a role to play in creative endeavors and making extra income.