AI is on everyone’s mind
Artificial intelligence goes beyond improving employee productivity, unleashing digital assistants, and forecasting inventory. Implemented fairly and equally, AI means more jobs and opportunities, not fewer, as it becomes more widespread. It means freeing up human effort beyond the mechanical and mundane tasks that have defined work since time immemorial.
The workplace lies at the core of AI’s potential, and its adoption could mean the difference between greater equality for workers around the world or whether income disparities continue to widen, says a new report on global AI equity published by the United Nations and the International Labor Organization. “AI has the potential to impact many aspects of our daily lives, but its impacts are likely to be most profound in the workplace,” the report says.
Moreover, AI will affect working conditions and the quality of work. There is a risk that AI will become too dominant, Berg and his co-authors say. This is caused by “the expansion of algorithmic control – a work environment in which human tasks are assigned, optimized, and evaluated through algorithms and tracking data.”
Algorithmic control is a trend that is driving digital labor platforms as well as industries such as the warehouse and logistics sectors, which “reduces workers’ autonomy to organize and set the pace of their own work, and workers have little ability to provide feedback or discuss the organization of their work with management.”
In the process, they say, “whether technology positively or negatively impacts working conditions will depend in large part on the voice that workers have in the technology’s design, implementation, and use. Such agency depends on opportunities for worker participation and dialogue.”
Economic growth will depend on organizations’ ability to better involve employees in decisions about AI adoption, say report authors Janine Berg, Mehdi Sunene and Lucia Velasco. “Whether the technology has a positive or negative effect on working conditions will depend in large part on the voice that employees have in the technology’s design, implementation and use.”
Opportunities for career advancement are emerging at several stages of the AI ”value chain” identified in the report. Types of AI-related opportunities include:
- Data collection: “Data is fundamental to the development and operation of AI systems. Human-prepared data is fed into AI systems to help them learn the connections and patterns they need to function. With global connectivity, data collection will continue to provide essential raw material for future AI applications.”
- Data curation and annotation: “The data collected is typically unstructured. Highly skilled data engineers preprocess the data into a usable format, but ‘data labelers’ are needed to label and classify the data so it is usable. “Workers are accessible through application programming interfaces, and programmers can invoke them with a few simple lines of code while working on an algorithm. While there are still many data labelers working in the United States and Europe, much of the work is taking place in developing countries due to the low compensation associated with the job.” Berg and his colleagues estimate that there are tens of millions of data labelers worldwide, and that this low-paid occupation is “likely to experience double-digit growth over the next five years.”
- Content Moderation: This is “the process of monitoring and filtering user-generated content on digital platforms, such as social media, forums, and websites, to ensure that it complies with the platform’s guidelines and policies. Content moderation can be performed manually by human moderators, or automatically using algorithms and machine learning tools. Even when algorithms and machine learning tools are used for content moderation, there is usually always a human involved in the process. While these technologies help automate and scale the moderation process, they are not perfect and may miss mistakes or nuances that a human moderator would be able to pick up on.”
- Model design, model training and tuning, deployment and maintenance: Unlike data annotation work or content moderation, these areas of AI work, which involve designing and building the necessary infrastructure, “require significant investments in research and development, as well as the skills of highly qualified computer scientists and graduates from other STEM fields.”
While low-wage jobs such as data collection, annotation, and content moderation are typically performed in developing regions of the world, more advanced tasks such as model design, training, and deployment create a large disparity in the economic benefits of AI.
“Disparities in access to digital infrastructure, advanced technologies, and quality education and training are deepening existing inequalities, especially as the global economy transitions toward AI-driven production and innovation,” the co-authors warn. “Developing countries risk being left behind, exacerbating economic and social disparities. They call for concerted efforts to bring AI development to disadvantaged regions by strengthening digital infrastructure, building AI skills, and ensuring quality jobs.”
As a result, “the AI gap is stark,” the co-authors write, “and such investments are costly, putting developing countries and their startups at a severe disadvantage. For example, OpenAI spent approximately $78 million in compute costs to train GPT-4, while Google’s Gemini Ultra is estimated to have cost $191 million.”
