"Ghost work" is an invisible term for anthropologist Mary L. Gray who supports the technology platform. When Gray, a senior researcher at Microsoft Research, arrived in the company, he learned that in order to build AI, people need to manage and organize their data to support training algorithms. "I basically started asking engineers and computer scientists around me who are paying to label the image and sorting tasks and clean up the database?" "Gray says. Some people said they do not know. Others said they did not want to know and worried that when they looked too closely, they could find unusual working conditions.
So Gray decided to learn herself. Who does what they need to do to run this platform, and who is often invisible? Why do they do this? Why would you leave? What are their working conditions?
Gray collaborated with fellow MSR senior researcher Siddharth Suri and wrote Ghost Work: Blocking Silicon Valley and building a new global underworld (Houghton Mifflin Harcourt).
The Verge told Gray about the findings and the implications for the future of employment.
This interview was lightly edited for clarity.
One obvious example of ghosting is labeling the data to be provided to an algorithm. Content mediation is another. What is another example?
Create surveys, subtitle translation tasks, and all sorts of enterprise services. Web research, location address verification, beta testing, user testing for user design. Everything you can think of as knowledge work such as content creation, editorial writing, and design work. Give me a name. The list is endless. All of this is an online deployment. It is everything we are used to seeing in the office. This is to dismantle into a full-time job and turn it into a project for countless people.
Basically, is it true that all technology companies depend on or depend on ghost work?
It would be difficult to find a business that sells AI that does not depend heavily on ghost work. I do not rely on creating a basic product or much of it today. There are many start-ups and businesses called "business insight" or "intelligence and analysis." It uses crowdsourcing or collective intelligence and relies on ghost work. There is no need for people to rummage unstructured data.
Sometimes people think that as technology improves, ghosting is no longer necessary. But you say, "The great paradox of AI is that the desire to eliminate human work creates new challenges for humans." So obviously you do not agree with that belief. Why not?
What specific tasks may change. Artificial intelligence means believing that people do not need to label data means that the language will not change. The style will never change. In particular, it is difficult for the service industry to fully automate because it is human ability to listen to someone's voice and register their quiet anger. So there are cases where AI always claims to be inadequate.
Engineers are always surprisingly optimistic about the opportunity. As an anthropologist, I know how complex it is to think culturally about these questions. What about those who want to translate to local and slang and translate language and transcode even if they reach 100% confidence in speaking English through the strength of the Midwest? Each time you see an automatic translation of a conversation, you can see where the language is damaged, and often the name of someone else.
The calculation that AI is difficult to capture because there is not enough data available to model the following statement using Spanglish. There is a problem with the above. . We've already easily automated everything effectively.
One interesting fact you mention is that there are no good labor statistics about how many people are doing ghost work. Why?
The biggest challenge is that the way in which a job is counted is related to a professional identity or to a clearly defined competency or skill, and no one is oriented to a project-based work world. We do not have a language to describe image tags or subtitles. One of our findings is that people really have different mental models. They may or may not be self-employed. When you create a content farm, it may or may not be classified as a journalist. Therefore, it may or may not be possible to answer survey questions to help measure this workforce. Needless to say, ghost work is scattered all over the world and there is no world-wide bureau of labor statistics.
The key question for this book is: Who are the ghosts? So, who are they? It seems to be almost everyone.
When we received the initial set of surveys on four different platforms we surveyed, there were many men and women, but it took a lot of time. People were educated in college, but it was not surprising because of their broad links to knowledge work and information services.
They are all of us. These are people who can not access the network to promote to full-time jobs for reasons of social capital. It is a pattern seen as sociological or anthropological. They are first generation universities. This is a group of people who do not have strong social ties to the elite.
What are the people's motives for this work?
There is only one type of motivation, not one type, to perform this task. There are people in the core group who go back to this task. Often due to different time constraints. People say they do not have time to commute and they say they are commuting from a relatively paid workplace for at least two hours, trying to reduce the money they can afford. It is the calculus they are making here. So they decide to do this effectively. Once you find a way to make enough money on a sufficient platform, you can meet your needs by collecting the equivalent of a full-time salary. We always call them "on" and are changing this to full time depending on the number of income streams. However, people in this group occupy a small percentage of 10-15% depending on the platform. This is what research is telling about all these platforms. People in the core group are doing a lot of work.
Then there are "regular customers" who have deep canals of people who can come in at any time. A regular customer is always making a person reunite. Because when "always on" goes out, there are people who can get in enough of the regular people's pile. They often become caregivers. There was another motivation; They pursued another passion project or returned to the education and training process and provided a means of funding.
Finally there is the long tail of the experimenter. Those who try one or two projects should find out that this is not the case and leave. The most important part of doing an anthropological work is meeting people we have left and knowing why. And I never felt that I was getting enough support from a fellow community that helped me save money, and it was too hard to figure out. And it was cognitively exhausted.
The hallmark of this kind of market is that anyone can work for others. What happens in that kind of environment?
People who are ordinary or "always on" get the same basic framework that they do through their investments. This is an amazing self-protection policy because workers invest in sending their work back to the pool. They want to see if their colleagues are doing well. If not, then you can be interested in getting the next job.
Enterprises need to invest equally in the responsibilities of the supply chain. If they are reliant on lowering their investment costs and are ready for what they need most and are willing to jump into the project, everyone has the opportunity to refresh and overtake the incoming person. Otherwise, it can not last in the labor market.
But the company does not. They do not create responsibility, trust, or culture to help ghost workers.
Whether you talk to any of these companies, most people think they get this automation and "they need these people for a while." This is our problem and our problem historically. After the Industrial Age: Heavily heal people who do accidental things that can not be automated. We do not pay attention to these people and the working environment, but we start to treat them as something that can eventually be replaced. It does not evaluate the fact that you are performing an operation that can not be performed by a mechanical process or a calculation process.
I hate parallelism with horsepower. This is different from turning a horse into a car. People do not do mechanical work. They are expanding something that is clear about humans, creativity and interpretation.
What should we do to solve this problem? What is the policy proposal?
At least that means making everyone's contribution important. The first step is to identify the contributors. In Bangladesh, there was a big change in the fabric when the company that sold the product had to talk about who made the shirt behind it. There should be a clear record of gratitude to those who have contributed to labor in their products or services. Consumers must be able to keep track of the supply chain of people who helped them achieve their goals.
This is about regulating employment types that are not suitable for full-time or part-time employment or that are not suitable for self-employment. I believe this is the moment when the classification of employment is no longer functioning. Everyone who has worked at the age of work must set the standards that the company provides.
To use contract work happily because companies need to constantly deliver new ideas and new aptitudes, the only way to make a difference for both sides of the enterprise is to let people jump into the pool. And people do that when they have health care and other provisions. Universal care, universal education for the public good. It will help all businesses.
I would like to be informed in many ways by people who explain working conditions. We do not describe operations for certain types. Describes today's conditions for project-based job-based operations. This can happen to everyone's job. I hate that it may be motivated because I have to be interested in everything because this is happening to many people. The message of this book to me is as follows. Let's make this sustainable and enjoyable without just managing it. Stop working to wrap our lives in the workplace and begin to serve our lives.