The bottom of the AI ​​pyramid: 'data annotator' earning $15 an hour

The application of artificial intelligence is already very popular. ChatGPT looks like "magic that responds to requests", but behind it is actually the credit of labor-intensive industries.

According to media reports such as CNBC and Gizmodo, OpenaAI has hired a large number of outsourced labor to assist them in completing "necessary data labeling tasks" - that is, labeling data. The more formal name is "data labeling" "member", "data annotator", or "AI trainer".

The so-called labeling means to put different labels (voice/picture/text, etc.) on the data samples to be analyzed by AI to help the AI ​​model better identify specific items in the data set and respond to user requests faster.

This is the most basic work of AI model training.

But this job is undoubtedly boring. It contains a lot of repetitive work. Operators only need to identify the type of data samples every day and then select different labels.

Alexej Savreux, a data annotator interviewed by CNBC, said:

We are workers, but without us there would be no artificial intelligence language system.

You can design all the neural networks you want, you can get all the researchers you want involved, but without tags, you don't have ChatGPT. You have nothing.

More importantly, The salary for such a job is $15 per hour-higher than the minimum wage in various states in the United States, but not much higher (Kansas City, where Savreux is located, Minimum wage $7.25).

Currently, domestic AI-related companies are also hiring for such positions.

The technology media "Whiplash" pointed out in an article in March that compared with the high salaries in the AI ​​industry, the salary of data annotators is not high.

"A picture costs 9 cents, and I can make 100 pictures a day." Lili said that if they are all qualified, they can earn 90 yuan a day.

"Different labels have different prices." He Wenxin said that his salary at the time was around 3,000. The monthly salary of basic data annotators is mostly between 2,000 and 4,000 yuan, but due to the speed and quality of annotation, "it is difficult to get the salary promised to you during the interview."

On some recruitment websites, Whipman searched for "data annotation" and set a salary range between 2,000 and 8,000 yuan. Some special annotations, such as minor languages, high-precision drawings, etc., will have higher salaries.

Artificial intelligence, focusing on artificial intelligence

Outsourcing repetitive tasks is not a phenomenon unique to the artificial intelligence industry.

CNBC pointed out that Silicon Valley has always relied on the labor of thousands of low-skilled, low-wage outsourced workers to build its computer empire, but these workers have always been in an "inferior" status:

These jobs are precarious, on-demand, and people are employed directly by a company through a written contract or through third-party providers who specialize in temporary work or outsourcing.

Benefits like health insurance are sparse or non-existent — meaning lower costs for tech companies — and the work is often anonymous, with all credit going to tech startup executives and researchers.

Now, the artificial intelligence industry is also following this rule of the game. With the booming development of artificial intelligence, more and more data annotators are being hired, but more and more of this basic labor is being ignored.

Sonam Jindal, project leader of the nonprofit organization "AI, labor and the economy at the Partnership on AI (PAI)" said:

Much of the discussion surrounding artificial intelligence is very welcome.

But we’re missing an important part of the story: This still relies heavily on a massive human workforce.

Tech giants in the whirlpool

Of course, as outsourcing positions such as data annotators have been seen by more and more organizations, this has also caused technology giants to face "trouble".

Earlier this year, Time magazine reported that OpenAI was relying on low-wage Kenyan outsourced labor to flag text containing hate speech or sexually abusive language so that its models could better identify “toxic” on their own. "content.

In Nairobi, Kenya, more than 150 people who have worked on AI for Facebook, TikTok and ChatGPT voted to form a union, citing low wages and the mental burden of the work.

Another media outlet Semafor reported in January this year that

OpenAI employs approximately 1,000 remote outsourced workers in places such as Eastern Europe and Latin America to label data or train company software on computer engineering tasks, where wages are As low as $2/hour.

Another data that contrasts with this is that,

As of January this year, OpenAI has approximately 375 employees.

And a spokesperson for the company said no one was available to answer questions about its use of artificial intelligence to outsource workers.

PAI warned in a 2021 report that demand for so-called "data-stuffing jobs" was surging, with the organization recommending the industry work on fair compensation and other improved practices and last year publishing a guide for companies Voluntary guidelines to follow.

CNBC pointed out that Google’s AI subsidiary DeepMind is the only technology company so far to publicly commit to complying with these guidelines.

Jindal said:

Many people have realized that this is a very important thing. The challenge now is getting companies to do it.

This is a new job created by artificial intelligence, and we have the potential to make this a high-quality job where the workers who do it are respected and valued for their contribution to making this progress possible.

The above is the detailed content of The bottom of the AI ​​pyramid: 'data annotator' earning $15 an hour. For more information, please follow other related articles on the PHP Chinese website!