Kansas City resident Alexej Savreux, 34, is among the unsung heroes working behind the scenes to teach AI systems, such as ChatGPT, how to analyze data for generating text and images that have impressed users worldwide. Savreux, along with other contractors, earns around $15 an hour without benefits for tasks such as labeling photos and predicting app-generated text to enhance AI accuracy.
These contract workers have devoted countless hours to refining OpenAI's systems, providing valuable training data in the form of sentences, labels, and other information. Although this work may not result in fame or fortune for individuals like Savreux, it remains crucial in the AI field, where the allure of groundbreaking technology often overshadows the contributions of gig workers.
"You can design all the neural networks you want, you can get all the researchers involved you want, but without labelers, you have no ChatGPT. You have nothing,"
- comments Alexej Savreux
Defined by its unsteady, on-demand nature, gig work typically involves employment through contracts written directly by the company or by a third-party vendor specializing in temporary or outsourced labor. Benefits, such as health insurance, are often scarce or absent, resulting in lower costs for tech companies. Furthermore, the work is usually anonymous, with credit attributed to tech startup executives and researchers.
In a 2021 report, the Partnership on AI predicted a surge in demand for "data enrichment work" and advised the industry to commit to fair compensation and improved practices. The organization published voluntary guidelines last year, but DeepMind, a Google AI subsidiary, is the only tech company to have publicly committed to them thus far.
As demand grows, AI contract workers are seeking better conditions. In Nairobi, Kenya, over 150 individuals who have worked on AI for Facebook, TikTok, and ChatGPT voted to form a union on Monday, citing low pay and the mental strain of the work, as reported by Time magazine. While Facebook and TikTok have not yet responded to requests for comment, OpenAI declined to comment on the matter.