Human Trainers: The Unsung Heroes Behind Sophisticated AI Models

Human trainers with specialized knowledge, including historians and PhDs, are increasingly essential for refining AI models like Cohere and ChatGPT. Companies like Invisible Tech, employing thousands of highly-educated trainers, partner with AI companies to mitigate errors known as hallucinations. This trend underscores a growing niche for well-paid experts.

In the initial development stages, AI models such as ChatGPT or its competitor Cohere required vast teams of low-cost workers to help distinguish basic facts like identifying a car versus a carrot in an image.

But as the race to create more sophisticated AI models intensifies, companies are now enlisting a rapidly expanding network of human trainers with specialized knowledge. These trainers are experts in fields like history and science, some even holding doctorate degrees. 'A year ago, we could get away with hiring undergraduates to generally teach AI how to improve,' said Cohere co-founder Ivan Zhang. 'Now we have licensed physicians teaching the models to navigate medical environments, or financial analysts guiding them in financial settings.' Cohere, valued at over $5 billion, collaborates with a startup called Invisible Tech for advanced training.

Invisible Tech, founded in 2015, employs thousands of trainers globally and has partnered with major AI firms like OpenAI and Microsoft. The startup's founder, Francis Pedraza, highlighted that their trainers, who can earn up to $200 per hour for high-expertise topics, help reduce AI errors known as hallucinations. Human trainers first entered the AI field through low-paying data-labelling tasks but now find a lucrative niche in shaping sophisticated AI models.

(With inputs from agencies.)