Indian gig workers toil at frontlines of AI revolution

🌈 Abstract

The article discusses the growing importance of data annotation services in India, particularly in the field of artificial intelligence (AI) model training. It highlights the increasing demand for skilled annotators, the evolution of annotation services, and the role of Indian workers in this industry.

🙋 Q&A

[01] Data Annotation Services in India

1. What is the current state of the data annotation services industry in India?

• India is emerging as a hub for data annotation services, with a large workforce of flexible workers, mid-tier business analysts, and skilled data engineers contributing to building high-quality datasets.
• According to TeamLease, a HR services company, there are 20,000 full-time workers engaged in the managed services paradigm as annotators in India, and 50,000 Indian annotators are actively employed as independent contractors across international platforms.
• The number of annotators worldwide is expected to double every three years, reaching almost 6 million workers in this field by 2040.

2. How are global AI companies leveraging India's data annotation services?

• Companies like Databricks, Fractal, Tredence, and startups like Cropin and Minus Zero are expanding their in-house teams of experts for faster and more cost-effective data annotation, while also relying on outsourced services in India.
• The need for annotation and the complexity of tasks have grown significantly for training large language models (LLMs), and India's skilled labor and lower operating costs make it an attractive destination for such services.

3. What are the key trends and challenges in the data annotation services industry?

• The original trend started around 2003-04 with e-commerce companies using workers in India to label products and create catalogues.
• While human annotation is crucial, self-supervised learning and the availability of open-source datasets are reducing the cost, time, and effort needed for manual data sorting and marking.
• As the complexity of training multimodal LLMs across text, speech, image, video, and code increases, especially in low-resource languages, skilled annotators will be required to build ethical guardrails into these innovations.

[02] Opportunities for Professionals in Data Annotation

1. How are professionals, including doctors and business analysts, engaging in data annotation services?

• Ikshita Nagar, a young Delhi doctor preparing for the PG entrance test, uses data annotation as a practice ground while preparing for her exam, seeing it as an opportunity beyond a second income.
• Annotation is evolving as a sub-segment at multiple firms, with a minimum qualification of a business analyst possessing domain knowledge.
• Practicing doctors can also participate in innovation happening in corporate healthcare through data annotation work.

2. What are the earning potential and skill requirements for specialized data annotators?

• While an average labeler can make Rs 25k-30k on platforms like Flexibench, a radiologist can make up to Rs 1 lakh per month for a few hours of work.
• Annotation for LLMs is more nuanced, requiring annotators to look at aspects like sentiment, in addition to the classical AI or ML tasks.
• Skilled annotators with specialized domain knowledge, such as in healthcare, are in high demand to build ethical guardrails into the development of complex AI models.