Artificial intelligence (AI) is rapidly transforming the global economy. Machines now deliver fast and accurate results. However, AI’s intelligence depends heavily on human labour behind the scenes. This work is mostly invisible and often outsourced to developing countries.
Human Role in AI Training
AI systems cannot understand raw data without human help. Data annotators label images, videos, audio and text to train AI and Machine Learning (ML) models. For example, large language models (LLMs) like ChatGPT need labelled data to recognise colours or objects. Self-driving cars use labelled video footage to differentiate traffic signs from pedestrians. The better the data quality, the better the AI output. Human labour is essential in creating these datasets.
Steps in Training Large Language Models
LLMs are trained in three steps – self-supervised learning, supervised learning and reinforcement learning. Initially, the AI learns from large internet datasets. Then, human annotators fine-tune this knowledge by labelling and giving feedback. This improves accuracy and removes errors over time. Human input is critical in the supervised and reinforcement stages.
Outsourcing and Working Conditions
Most data labelling work is outsourced by Silicon Valley companies to workers in countries like Kenya, Pakistan, China and the Philippines. These workers often earn very low wages and work long hours. Some tasks require specialised knowledge but are assigned to non-experts, causing errors in AI outputs. For instance, medical scan labelling without proper expertise has led to mistakes. Recently, companies have begun hiring experts for niche tasks.
Invisible Labour in ‘Automated’ Features
Even AI features marketed as fully automated rely on human labour. Social media content moderation depends on human moderators labelling harmful material. Exposure to graphic content causes mental health problems among these workers. AI-generated audio and video also require human actors for training. Some children have been engaged for such tasks. This human work remains largely unrecognised.
Labour Exploitation and Lack of Transparency
In 2024, Kenyan AI workers wrote to US authorities denoting poor working conditions. They described low pay, long hours and exposure to disturbing content. Complaints often lead to dismissal and union suppression. Most workers do not know the end company they serve as work is subcontracted through digital platforms. These platforms pay per microtask and enforce strict surveillance. This fragmented labour network lacks transparency and encourages exploitation.
Need for Regulation and Fair Labour Practices
The AI industry depends on ghost workers whose contributions go unacknowledged. Informal labour arrangements help companies cut costs but harm workers. There is a pressing need for laws regulating AI companies and digital platforms. Such rules should ensure transparency, fair wages and dignity for all workers involved in AI development.
Questions for UPSC:
- Critically analyse the impact of globalisation on labour rights in the digital economy with suitable examples.
- Explain the role of human labour in the development of artificial intelligence and discuss the ethical implications involved.
- What are the challenges faced by workers in informal sectors? How can government policies ensure their protection and welfare?
- Comment on the need for regulatory frameworks in emerging technologies like AI and digital platforms. What measures can be adopted to balance innovation and labour rights?
