Meet the working poor who supply the ChatGPT – 08/03/2023 – Tech

Since it was launched at the end of 2022, the ChatGPT artificial intelligence program has aroused admiration for the technological advancement it represents — but also fears about its future impacts.

Faced with this chatbot capable of answering almost all user questions and producing texts that seem to have been written by a human, questions arose such as: will students use it to do their homework? Politicians to write their speeches?

Was this article you are reading written by a human or a robot?

This type of program has also raised concerns about works that will cease to exist due to automation and with copyright, since these tools obtain information from the Internet and generally do not cite sources.

But there is another controversy so far little talked about: it has to do with the hundreds of thousands of workers, many of them low-income, without whom AI (artificial intelligence) systems like ChatGPT would not exist.

We are talking about the “hidden workforce”, as the non-profit organization PAI (Partnership on AI) called it, which brings together representatives from universities, civil society organizations, the media and the industry itself involved with AI.

This hidden force is made up of people subcontracted by large technology companies, often in poor countries in the southern hemisphere, to “train” AI systems.

‘Labelers’

These men and women perform a tedious task—and potentially harmful to mental health, as we’ll get to later—but one that is essential for programs like ChatGPT to work.

They label millions of data and images to teach the AI ​​how to act.

Take, for example, the chatbot of the moment.

When you ask ChatGPT a question, the program uses about 175 billion “parameters” or variables to decide what to answer.

As already mentioned, this AI system uses information obtained from the internet as its main source. But how to distinguish the contents? Thanks to references “taught” by humans.

“There is nothing intelligent about artificial intelligence. It has to learn as it is trained”, explains Enrique García, co-founder and manager of DignifAI, an American company based in Colombia.

The company hires these data “taggers”.

In the technology industry, this type of activity is called “data enrichment”.

Ironically, despite being essential work for the development of AI, data enrichment is the poorest link in the production chain of large technology companies.

A fact that has been recognized by the Partnership on AI organization.

“Despite the critical role these data enrichment professionals play, a growing body of research reveals the poor working conditions these workers face,” said the organization, of which OpenAI, the company that created ChatGPT, is a part.

Less than $2 an hour

An investigation by Time magazine revealed that many of the third-party labelers OpenAI hired to train its ChatGPT were paid between $1.32 and $2 per hour.

According to a report by journalist Billy Perrigo, the technology company, which has Microsoft among its main investors, outsourced the data enrichment work through a company called Sama, based in San Francisco —which in turn hired workers in Kenya. for the activity.

Through a statement, an OpenAI spokesperson said that the outsourced company was responsible for managing the wages and working conditions of the taggers hired to work on ChatGPT.

“Our mission is to ensure that AI benefits all of humanity, and we work hard to build safe and useful AI systems that limit bias and harmful content,” the spokesperson said.

Sama also hires for Google and Meta labelers in other low-income countries such as Uganda and India. The company bills itself as an “ethical AI” and claims to have lifted more than 50,000 people out of poverty.

However, Martha Dark, director of the British activist organization Foxglove —whose objective is “to take on technology giants and governments, for a future where technology is used to benefit everyone, not just the rich and powerful”—, reckons that companies tech companies use outsourcing to pay workers much less than they should.

“All of these companies are multi-billion dollar and it is frankly inappropriate that they are paying $2 an hour to the people who make these platforms possible,” he said.

But for DignifAI’s Enrique García, the wage controversy “is a matter of perspective.”

In Europe and the United States, it can be understood that earning that amount is not enough, but in other parts of the world, this can be a good salary, he argues.

“Many people criticize our sector for the salary issue, but at DignifAI our minimum salary is US$ 2.30 (R$ 12) an hour, which represents 1.8 times the minimum salary in Colombia”, he says.

“If the project is more complex and requires specialized professionals, such as architects or doctors, the salary can reach US$ 25 (R$ 130) per hour”, says García.

Although he recognizes that there are companies that pay below the minimum wage, the businessman considers it unfair to focus only on this sector.

“There are outsourcing dynamics in many industries, not just this one, so it’s not fair to label us ‘digital exploitation’ either,” he says.

social impact

García also points out that there are several companies in the sector, like yours, that have a social impact and the objective of “increasing people’s productivity and dignity”.

DignifAI’s motto is “outsourcing dignity through artificial intelligence”.

The company is based in Cúcuta, on the border between Colombia and Venezuela, and seeks to provide work for Venezuelan migrants and Colombians who have migrated internally.

“Many of them, before working with us, earned US$4 or US$5 a day. For this vulnerable population with no job market options, earning 1.8 times the Colombian minimum wage is quite attractive,” he says.

Ingrid, 42, a Venezuelan who arrived in Colombia at the end of 2018, confirms this.

With a degree in pedagogy, Ingrid, who preferred not to give her last name, told the BBC that she currently cannot teach because she has not yet validated her diploma in Colombia.

She claims that working as a labeller for DignifAI has allowed her to earn a living and also prepare for another profession.

“I work four hours a day and have managed to use the remaining time by taking a design course”, says the pedagogue.

Despite no longer working as a data tagger, having been promoted to the position of project supervisor, she does not hesitate to recommend this job.

“It’s more rewarding, less tiring and better paid than being a waitress, part-time assistant or doing physical work,” she says, adding that most of her colleagues are housewives, street vendors or students.

Mental health

In addition to salary, another issue regarding the working conditions of data taggers is the effect on mental health.

It’s not the tedium of the task that most worries some experts —although that’s another criticism leveled at this work—but the toxic material to which some of them are exposed.

One of these workers’ jobs is to teach the AI ​​program what information is not suitable for publication – delving into the darkest corners of the internet and labeling the violent, sinister and perverse material there, in order to tell the machine to ignore it all.

According to Martha Dark of the organization Foxglove, doing this work “can cause post-traumatic stress and other mental health problems for many workers.”

His organization assists a former Sama employee who worked as a Facebook moderator in Kenya. In 2022, he sued Sama and Meta, owner of the social network, for the psychological damage he suffered. The case is still pending in the Nairobi court.

“These jobs come at a cost to the mental health of those doing them and they should provide adequate psychiatric treatment as well as a fairer wage,” Dark told the BBC.

According to the activist, the big technology companies have a lot of financial resources to provide this type of assistance, but they don’t do it because “they put profit above the safety of their workers”.

Enrique García acknowledges that large companies could invest more in hiring taggers, but claims that demanding too much from them could lead them to look for workers elsewhere.

“It may be that big tech can pay more, but we’re very grateful for the opportunities,” he says.

“At least we are bringing income generation opportunities here where, without this alternative, they would not exist.”

This text was published here.