Summary of the Reading
The authors talk about a hidden or invisible power named as a Ghost Work that work side by side with software and AI systems to make sure that data is accurate when the AI system fails to do so. Users of systems like google search, YouTube, Facebook, and other applications don’t know that there are people working behind the sense to make sure that the internet is safe. There are thousands of workers that work on demand on a specific tasks in a form of projects that eventually either built a product or help to make sure that the software is working correctly. The author called these workers a shadow workforce and their work as a Ghost Work. The author suggesting that 38% of the current work will turn into ghost work in the future because of the advancement in technology and automation. Companies like Amazon are assigning people tasks to people to manually inspect some content like hate speech in Twitter or inappropriate pictures. Such tasks can’t be automatically inspected by AI systems. The human being in the loop when the AI can’t do the perfect job. The way workers can get tasks is competitive. The workers must claim their tasks using a request dashboard and it is first in first serve. The mix between human and computer computation is called crowdsourcing, microwork, crowdwork, and human computation. The ghost work doesn’t follow employment laws, or any government regulation and it is cross boarders and the countries how do the work more are US and India because of English language fluency. Work can be micro tasks like deterring if a post on Facebook is a hate speech or macro work like ImageNET project that labeled millions of pictures to be used in AI systems. Ghost work is also known as on-demand gigs economy and there are 2.5 million adults in the United States that participate in this kind of work. Amazon was one of the first companies that depends of ghost work and have built MTruck software handle its work load to give work to thousands of people to correct spelling or typos or any incorrect information. Ghost work is important to label data to make better prediction for algorithms like machine learning, machine translation, and pattern recognition. As automation advances people might start losing their jobs but the ghost work will be more requested and that can balance out the need for the workforce.
Reflections and Connections
What I found interesting about the reading was the way people around the world are working together on tasks that eventually produce a product or service and to make the internet safe. My understanding was companies uses AI systems to flag content and then employees or contractors but didn’t know that people across the globe are working together on task. I also found interesting that the interaction between human and the AI systems will continue which can decrease the concerns that the machine will take over people job in the future.
I agree with what was mentioned in the paper regarding the fusion of the code and human and the future of the workforce as they will turn more to invisible or ghost work.
As GitHub acquired by Microsoft, I think the software development will also follow the same patterns by assigning tasks to software developers across the globe to build software modules that will participate in building large scale applications.
Questions
- What things we need to put in our consideration while doing research in the AI area that might affect human in the loop?
- How do we build software systems that can automatically take the feedback of manual inspection or tagging and adjust its behavior for similar incidents?
- What is our role as graduate students or researchers to shape the future of the ghost work?
