Microsoft recently announced that ChatGPT would be added to its Azure services and made it generally available so that users could access a variety of AI capabilities. In the midst of the investments and growth, a recent Time article uncovered the shadow side of this popular AI chatbot, outlining the realities that underlie the creation of this AI model.
Sama and OpenAI Working Together
In order to discover and categorise sensitive photos and text data that is later used to train ChatGPT so it can spout out stunning responses devoid of toxicity, OpenAI stated that it used Kenyan labour in the development of a tool that classifies problematic information.
Sama, an AI company based in San Francisco, worked with OpenAI to create this chatbot; the two organisations parted ways in February 2022. According to sources, the former made this choice in response to OpenAI’s purported demand for unlawful content, such as child sexual abuse, rape, and other acts of violence, in order to train an AI project unconnected to ChatGPT.
$2 per hour was paid to employees to identify sensitive content
Data laborers were reportedly contracted out under Sama’s contract to Kenya, according to new information. The outsourced labelers examined the content, which included sexual abuse, hate speech, and general violence, for which they were paid between $1.32 and $2 per hour. Additionally, some data laborers told Time that they did not receive clear instructions on how to classify the content they examined.
According to the study, one was asked to read a graphic scenario in which Batman’s sidekick, Robin, is raped and was unsure whether to classify it as sexual violence because Robin ultimately engaged in sexual activity.
Sama’s decision to stop showing such material
Aside from its dispute with Sama last year, OpenAI also offered its graphic content services to Microsoft and Google. According to unnamed sources, Sama has now chosen to discontinue all of its operations by March 2023, including a $3.9 million deal with Facebook.
Our focus will be on computer vision data annotation solutions, signalling an end to all efforts on natural language processing and content control, according to a statement from Sama.
Upon its debut in November of last year, ChatGPT was heralded as one of 2022’s most stunning technological advancements. The sophisticated artificial intelligence (AI) chatbot can produce writing on virtually any subject or theme, from a Shakespearean sonnet reinvented in the manner of Megan Thee Stallion to intricate mathematical theorems explained in terms that a five-year-old may comprehend. It attracted more than a million users in a week.