After having her second child, Chelsea Becker took a year off without pay from her full-time job as a flight attendant. After she saw a video on TikTok, she found a side hustle training artificial intelligence models for her website called Data Annotation Tech.
Becker, 33, of Schwenksville, Pennsylvania, spends several hours each day sitting in front of his laptop and interacting with an AI-powered chatbot. For each hour of her work, she was paid a salary of $20 to $40. From December to March, she earned more than $10,000.
The boom in AI technology has made a type of gig work that doesn't require you to leave your home more sophisticated. The growth of large-scale language models, such as the technology behind OpenAI's ChatGPT, has increased the need for trainers like Becker who can speak fluent English and produce high-quality texts.
It's no secret that AI models learn from humans. Makers of AI systems like Google and OpenAI have long relied on low-wage workers, typically contractors hired through other companies, to help computers visually identify objects. (The New York Times sued OpenAI and its partner Microsoft for alleged copyright infringement.) Images may be identified.
But as AI technology has become more sophisticated, so has the job of those who must painstakingly teach it. Yesterday's photo tagger is today's essay writer.
These trainers typically have two types of tasks. One is supervised learning, where the AI learns from texts written by humans, and the other is reinforcement learning, where the chatbot learns from human feedback and evaluates human responses.
Companies specializing in data curation, such as San Francisco-based startups Scale AI and Surge AI, hire contractors and sell training data to large developers. Developers of AI models, such as Toronto-based startup Cohere, are also hiring in-house data annotators.
Researchers say the total number of these gig workers is difficult to estimate. But Scale AI, which employs contractors through its subsidiaries Remotasks and Outlier, said it typically had tens of thousands of people working on the platform at a time.
But like other types of gig work, the ease of flexible work hours comes with its own challenges. Some workers said they had never interacted with the managers behind the recruitment sites, and others said they were cut off from their jobs without any explanation. Researchers have also raised concerns about the lack of standards, as employees are typically not trained on what is considered an appropriate response for a chatbot.
To become part of these contractors, workers must pass a screening that includes questions such as whether and why their social media posts should be considered hateful. The other one requires a more creative approach and asks me to write a fictional short story about a green dancing octopus set in Sam Bankman Freed's FTX office on November 8, 2022. (On that day, according to FTX competitor Binance). (After acquiring Mr. Bankman Fried's company, they would soon withdraw from the transaction).
In some cases, companies may be looking for experts in their field. Scale AI has job openings for contract writers with master's or doctoral degrees in Hindi and Japanese. Outlier has job postings with requirements such as degrees in mathematics, chemistry, and physics.
“What makes AI really useful to users is the human data layer, and it needs to be done by really smart humans, skilled humans, and humans with specific expertise and creative talent.” Vice President Willow Primack said. An overview of data operations in Scale AI. “As a result, we have focused specifically on contractors within North America.”
Self-published novelist Arinzia Fenske had never interacted with an AI chatbot until she heard many stories from fellow writers who saw AI as a threat. So when she found a video on TikTok about her data annotating her technology, part of her motivation was to learn as much as she could about AI and whether the fears surrounding AI were justified. I had to see it with my own eyes.
Mr. Fenske, 28, who lives in Oakley, Wisconsin, said, “After actually working on it, I got a completely different perspective.It's reassuring to know that there's a human being behind it.'' Since February, she has aimed to commit to 15 hours of data annotation work each week so that she can support herself while pursuing a writing career.
Ese Agboh, 28, a master's student in computer science at the University of Arkansas, was given a job on a coding project that paid $40 to $45 an hour. She asks the chatbot to design a motion sensor program that will allow gym patrons to count their repetitions, and the computer evaluates her code written by the AI. In another case, she loads a dataset about groceries into the program and asks the chatbot to design it. monthly budget. Sometimes she also evaluates other annotators' code, which experts say is used to ensure data quality.
She earned $2,500. However, her account was permanently suspended by her platform for violating its code of conduct. Although she received no explanation, she suspected it was because the site only sought workers based in certain countries, so she was working while in Nigeria.
That's the fundamental challenge of online gig work. It can disappear at any time. With no one to turn to for help, frustrated contractors turned to social media, sharing their experiences on Reddit and TikTok. Jackie Mitchell, 26, has amassed a large following on TikTok thanks to her content about her side hustle, such as data annotation work.
“I understand the appeal,” she said, noting that side hustles are “unfortunately necessary” in this economy and “a hallmark of my generation and older generations.”
Public records show Surge AI owns the data annotation technology. Neither the company nor CEO Edwin Chen responded to requests for comment.
It is common for companies to hire contractors through subsidiaries. James Muldoon, a management professor at the University of Essex who focuses on AI data work, said these measures were to protect the identity of customers and avoid bad publicity associated with working conditions for low-paid contract workers. Said it was helpful.
The majority of data workers today rely on wages from gig work. Milagros Miceli, a sociologist and computer scientist who studies working conditions in data work, said that while “gamification means a lot of people are doing this for fun,'' the work Parts are still “done by workers,” he said. These are people who actually really need money and are using this as their main income. ”
Researchers are also concerned about the lack of safety standards in data labeling. Workers are sometimes asked to address sensitive issues, such as whether certain events or actions should be considered genocide or which gender should appear in an AI-generated image of a soccer team, but they are not trained on how to conduct that assessment.
“It's generally not a good idea to outsource or crowdsource safety or ethical concerns,” Professor Muldoon says. “You have to follow the principles and values and what the company actually thinks is the right thing to do on a particular issue.”