Artificial Intelligence (AI) models like ChatGPT and Cohere have evolved significantly over the years, requiring a vast network of human trainers with specialized knowledge to train the models effectively. In the past, basic facts like distinguishing between a car and a carrot were taught by undergraduates, but now, licensed professionals such as physicians, financial analysts, and accountants are involved in training these AI models. Cohere, a major competitor of OpenAI, collaborates with a startup called Invisible Tech, which employs thousands of trainers worldwide, offering competitive rates of up to $40 per hour depending on expertise.
Invisible Tech, founded in 2015 as a workflow automation company, shifted its focus to AI training after OpenAI approached them for help with reducing errors or “hallucinations” in their models. One of the challenges AI companies face is the generation of false information by AI models, leading to inaccurate outputs or hallucinations. By utilizing human trainers to differentiate between fact and fiction, AI companies aim to enhance the quality and reliability of their AI models. Invisible has become a primary training partner for genAI companies like Cohere, AI21, and potentially Microsoft.
OpenAI’s Human Data Team collaborates with AI trainers from companies like Invisible to gather specialized data for training models like ChatGPT. Various experiments are conducted to enhance the AI models’ performance, such as reducing hallucinations or improving writing style. By hiring workers with relevant degrees for specific projects, the burden of managing trainers is reduced for AI companies. Invisible has engaged over 1,000 contract workers to cater to OpenAI’s training needs alone, showcasing the growing demand for specialized trainers in the AI industry.
Competition in the AI training sector is increasing, with companies like Scale AI offering training data sets and AI trainers. Scale AI, valued at $14 billion, counts OpenAI as a customer but did not respond to requests for an interview. Invisible, a profitable company since 2021, has raised $8 million in primary capital and remains majority-owned by its team. As the demand for specialized AI trainers continues to rise, workers across various subjects are entering the field without the need for coding skills, creating a lucrative niche market for AI training services.
The growing demand from AI companies has led to the emergence of new firms offering similar services in the AI training space. Professionals from diverse backgrounds, including historians, scientists, and more, are becoming AI trainers, contributing to the development of advanced AI models across multiple languages. With the increasing complexity of AI models, the role of human trainers in teaching AI concepts like data analysis and model behavior is critical. The partnership between human trainers and AI companies is essential for enhancing the accuracy and performance of AI models in various industries.
