In 2021, Richard Mathenge of Nairobi, Kenya believed he had achieved his dream job when he was hired to train OpenAI’s GPT model. After years of working in customer service, he now had a meaningful opportunity that held a promising future. However, the work came with tremendous sacrifice; for nine hours a day, five days a week, Mathenge and his team were confronted with explicit content in order to teach the artificial intelligence how to filter such content away from users. Despite the traumatic nature of this job, the work Mathenge and his team did proved to be essential in producing advanced and sophisticated technologies, such as chatbots like ChatGPT and Google’s Bard.
The process of labeling and categorizing this explicit content includes labelling it as child sexual abuse material, erotic sexual material, and other types of illegal content. Many of the texts Mathenge and his team had to view were appalling, such as passages that described a father having sexual relations with an animal in front of a child. This type of work, known as Reinforcement Learning from Human Feedback (RLHF), teaches the AI models which values to prioritize based on human feedback.
Mathenge had a degree from Nairobi’s Africa Nazarene University in 2018 and quickly found a place in the technology sector. Prior to joining OpenAI, Mathenge worked with Sama, an AI annotation service which worked for OpenAI. After completing a project in LiDAR annotating for self driving cars, Mathenge was allocated to work with OpenAI’s model. Despite being promised $12.50 per hour, Mathenge and his colleagues were only paid about $1 per hour,in some times even less.
Mathenge’s colleague Mophat Okinyi still suffers the consequences of what he endured; the excessive viewing of explicit material resulted in severe anxiety, depression and insomnia, so much so that Okinyi’s wife eventually left him.
OpenAI stated that it made sure its contractors were provided mental health support, although Mathenge and Okinyi felt it was inadequate. After OpenAI contacted Sama for more information regarding working conditions, Sama decided to leave the content moderation space.
Despite the horrific cost of the experience, Mathenge and his team still take pride in their contribution to the success of OpenAI’s technology, which, as a result, now refuses to produce explicit scenes and advises users on potentially illegal sexual acts.
OpenAI is a business focused on artificial intelligence, with investors such as Microsoft, and it is widely considered a leader in the field. OpenAI has earned praise for its technical achievements and overall objectives. OpenAI’s CEO, Sam Altman, also serves as a mentor and advisor to Y Combinator, a startup accelerator.
Richard Mathenge has made incredible strides in his high tech career, earning a degree and finding a place in the technology sector as a contractor. Mathenge worked with Sama, then transferred to OpenAI and encountered the explicit content he was tasked with labelling. He has faced challenging situations to make a better livelihood and address poverty while working in a developing country. His dedication to OpenAI’s work has paid off and he is grateful for the help it provides and can only imagine its potential.
