As AI models like ChatGPT and Cohere have evolved, the sophistication of their responses has improved remarkably. While earlier versions of AI relied on large teams of low-wage workers to help models distinguish between simple facts, like identifying whether an image was of a car or a carrot, today’s more advanced iterations require a different caliber of trainer.
Gone are the days when undergraduate students could adequately assist in training AI models. The rise of generative AI has spurred the need for experts with specialized knowledge across various fields, ranging from medicine to finance. Ivan Zhang, co-founder of Cohere, pointed out that “a year ago, we could get away with hiring undergraduates. Now, we have licensed physicians teaching the models how to behave in medical environments, or financial analysts or accountants.”
Cohere, a business-focused AI platform valued at over $5 billion, partners with a company called Invisible Tech to meet this growing demand for high-caliber human trainers. Invisible Tech employs thousands of remote trainers worldwide to help improve AI models by reducing errors, commonly referred to as “hallucinations” in the AI world.
According to Francis Pedraza, founder of Invisible, the company has 5,000 trainers in over 100 countries, including PhDs and Master’s degree holders. These human trainers teach AI models to produce more reliable responses, especially in specialized fields like quantum physics, where hourly rates for trainers can soar to $200. Invisible has become a major partner for AI companies, including AI21, Microsoft, and OpenAI, the creator of ChatGPT.
Invisible’s role in the AI ecosystem skyrocketed in 2022 when OpenAI approached them with a specific challenge. Early versions of ChatGPT were prone to hallucinating, generating false information despite being based on historical data. Invisible provided OpenAI with human feedback to fine-tune the model, reducing errors and improving reliability.
While companies like Invisible and Cohere rely heavily on these human trainers, they are not alone in this space. Scale AI, another private start-up valued at $14 billion, also offers similar services, counting OpenAI among its clients. The growing demand for human trainers has led to the emergence of more firms offering specialized services to AI companies. Zhang of Cohere noted that his inbox is constantly flooded with messages from new firms looking to enter this niche industry.
This surge in demand is creating a lucrative new market for workers from a wide variety of fields, even those without coding experience. Whether it’s Swedish history, biology, or chemistry, human expertise is critical in ensuring AI models can provide accurate and contextually appropriate responses.
As AI continues to advance, the role of human trainers will only grow in importance. What began as a low-cost data-labeling job has now transformed into a high-paying, specialized field. The result? Smarter AI models capable of answering complex queries with increased accuracy thanks, in no small part, to the experts behind the scenes.