Shulou(Shulou.com)11/24 Report--

According to people familiar with the matter and internal communication documents, more than 30 dispatched workers who helped train the language model behind the hot chat robot ChatGPT were fired in March this year, according to news on April 23.

Screenshots of Slack's internal chat show that as of March 16, San Francisco-based outsourcing company Invisible Technologies had laid off 31 workers. However, OpenAI continues to hire across its company.

The screenshot also shows hundreds of Invisible Technologies workers known as "advanced artificial intelligence data trainers" working with OpenAI to help them train their GPT chat robots. According to one dispatcher, the company's artificial intelligence data trainers are responsible for improving the coding skills of models, enhancing their creative writing skills, or training them to refuse to respond to certain topics. The dispatcher spoke on condition of anonymity because he signed a confidentiality agreement, but insiders confirmed his identity and employment.

Camron Parritzban (Kamron Palizban), vice president of operations at Invisible Technologies, talked about layoffs at a staff meeting in March. In a leaked recording of the meeting, he said OpenAI wanted to reduce the number of dispatched workers because of changes in business needs. Palizban also said at the meeting that many of the projects that laid off dispatched workers did not provide a high enough return on investment for OpenAI.

OpenAI has slashed the number of dispatched workers. Invisible Technologies's relationship with OpenAI offers a glimpse into the data training of the ChatGPT maker. To a large extent, OpenAI has always kept these exercises secret.

Before the contract adjustment between OpenAI and Invisible Technologies, it was reported that the former had increased the number of employees for six consecutive months. As of January, OpenAI had employed nearly 1000 data tagging dispatchers in places such as eastern Europe and Latin America, according to people familiar with the matter.

The layoffs at Invisible Technologies come just two months after Microsoft injected $10 billion into OpenAI. But Invisible Technologies is not the only outsourcing company working with OpenAI.

Sama, an outsourcing company also based in San Francisco, terminated its partnership with OpenAI in February 2022 after it learned that its Kenyan employees were reviewing harmful content such as sexual abuse, hate speech and violence, according to a survey by time magazine.

In a statement to time, an OpenAI spokesman explained: "classifying and filtering harmful text and images is a necessary step in minimizing the amount of violent and pornographic content contained in training data and helps to create tools that can detect harmful content."

The work of artificial intelligence trainers according to Invisible Technologies dispatchers, the most basic duties of artificial intelligence trainers include reviewing conversations between artificial intelligence and its users to identify potentially illegal, invading privacy, offensive or error messages. The dispatched workers interviewed described their daily lives as follows:

After starting the shift, they first open the internal work browser to view the team's to-do list. They may click on the task "having a conversation on a random topic while browsing is prohibited," and then enter a query in the message box.

After submitting the query, OpenAI's model generates four responses. Dispatchers evaluate each response by opening the drop-down menu and selecting the type of error that exists, such as factual error, spelling or grammatical error, or harassment. The dispatcher then rated the severity of the error on a scale of one to seven, with seven representing a "basically perfect" answer.

Next, the dispatcher must carefully design a perfect response and submit it to show that the task is completed. The dispatcher said the results would be sent to quality inspectors at OpenAI and Invisible Technologies. This process needs to be repeated over and over again for each of the following tasks.

Camron Palizban, vice president of Invisible Technologies operations, said of OpenAI at the meeting: "they are at a stage where they are about to have a clearer understanding of where they are going."

Grace Grace Matelich, partner and operations manager at Invisible Technologies, said at the recorded meeting that the company had fired underperforming dispatchers based on performance indicators such as "quality" and "throughput" to complete the task.

Martledge said underperforming dispatched workers and those who were hired but did not "meet the certification threshold" were fired, although some were allowed to choose to move to different OpenAI teams. He added: "if you are still here today, I want you to know that this is because I believe in your ability to do a good job."

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.