xAI plans to hire thousands of people this year to help train its chatbot Grok, three employees told Business Insider, as it looks to gain an edge in the AI arms race with OpenAI, Meta, and Google.
Elon Musk's artificial intelligence company has five job listings for data annotation roles, seeking bilingual workers and legal and STEM specialists. The employees said they were told by managers about the hiring ramp-up.
xAI employs more than 900 data annotators, which it calls "AI tutors," according to a tally of the company's internal Slack. It brought in hundreds of tutors in preparation for the November presidential election, according to six current and former employees. Business Insider first reported that xAI was ramping up hiring in October of last year.
The workers asked to remain anonymous in order to avoid professional reprisal.
Musk has repeatedly emphasized the potential of AI, and he has used his startup to wade into the industry's war over funding, talent, and power. The company last year built what Musk has called the world's largest supercomputer in Memphis, Tennessee, and it launched Grok as a stand-alone app in January. xAI said during its latest funding round that the company had raised $6 billion.
On Monday, Musk and a group of investors submitted a $97.4 billion bid to buy the nonprofit that controls OpenAI; CEO Sam Altman said the company was "not for sale."
Representatives for xAI did not respond to a request for comment.
Inside the day-to-day work of an xAI tutor
Data annotators play a crucial role in developing large language models. By labeling, categorizing, and contextualizing raw data, they teach chatbots like Grok to understand the world.
The tutors have worked on various projects at xAI, including improving Grok's ability to generate images. To help it understand voice commands, they record themselves in noisy environments and analyze the chatbot's transcriptions. Three workers said the company was working toward expanding the pool of staffers who can train Grok in different languages.
Additionally, four workers said tutors were focused on teaching Grok to understand, summarize, and provide context to posts and images on X. (xAI rolled out that update in December.)
"Sometimes projects take months; sometimes it's only a few days and then you're getting a message from a team lead moving you to a new project," one worker said. "Sometimes it's a drop-everything situation."
At xAI, the tutors operate on six-month contracts as full-time hourly employees, according to job postings and conversations with workers. They work in tandem with engineers, but the data annotators use a separate Slack, called TeachX, and have different email domains, six current and former employees said.
Within TeachX, AI tutors are broken into several specialized groups, including STEM, coding, and translation, as well as a much larger group of general AI tutors.
The job listings indicate the workers can make $35 to $65 an hour and can be promoted to team-lead roles in a matter of months.
Six current and former workers said that while most tutors work from home, they're supervised using a system called Starfleet Academy that predicts the amount of time workers should spend on each task, monitors their screen time, and allows quality-assurance workers to rate their work. Three employees said workers typically take three to 10 minutes on a given task.
Four workers said that the tutors could be asked to justify their choices and could get marked down if a quality-assurance worker disagrees with them.
"Mainly Starfleet is a way to make sure we're not moving too fast," one worker said. "Sometimes it can become monotonous or you can get numb to flipping between prompts."
Data annotators are crucial to the tech industry
The AI tutors appear to make up most of xAI's workforce. A review of LinkedIn profiles suggests the company has about 100 noncontract employees, including engineers and technical staffers.
Unlike most AI companies, which often use third-party employment agencies, xAI hires its US tutors directly. Otto Kässi, a labor economist at the University of Oxford, said this model allows the company to have more control over the tutors' work and could offer greater data security.
xAI is far from the only company leaning on an army of data annotators.
Tesla, Musk's electric vehicle manufacturer, employs hundreds of in-house data annotators to train its Full Self-Driving software. Companies like Outlier.ai, Scale AI, SuperAnnotate, and Encord also provide annotators for Big Tech companies including Meta and Nvidia. OpenAI, which Musk helped found before resigning from the board in 2018, has used third-party agencies that can hire workers from developing countries.
"The most important thing is the quantity of data," Stefano Ermon, a generative AI expert from Stanford University, told BI. "Then you just need enough people to annotate that data."
Do you work for xAI or one of Musk's companies? Reach out to the reporter via a nonwork email and device at gkay@bjinnox.com or through the encrypted messaging platform Signal at 248-894-6012.
