The Turing Artificial Intelligence Data startup, one of a growing number of companies that provide human instructors for AI laboratories, said on Tuesday that their revenue tripled to $ 300 million last year and reached profitability.
Turing, based in Palo Alto, which claims to have among clients OpenAi, Google, Anthropic and Meta, was last valued at $ 1.1 billion by 2021.
As IA models became more sophisticated, the demand for human instructors with expert knowledge increased, boosting the assessment of startups such as Turing competitor, Scale AI, which was valued at $ 14 billion last year.
Based on what AI companies want as an improvement for their AI models, data companies find workers with specialized knowledge relevant to these projects, reducing the burden of managing hundreds of instructors by the developers of these models.
Turing claims to have access to over 4 million human experts, such as software developers or doctoral scientists, who can be hired to label data to AI models.
The costs are not cheap: a complex annotation can cost hundreds of dollars, Turing said, and advanced AI models may require millions of notes. For example, the goal used more than 10 million human notes by training the Llama 3 models, said Joe Spisak, Meta’s executive last year.
Continues after advertising
As AI laboratories reach the “data wall”, a term for when model performance reaches a level due to the lack of more internet training data, laboratories will increasingly depend on human data companies to make His smarter AI models, Turing’s executive president told Jonathan Siddharth to Reuters.
“Companies like Turing are helping to compensate for the data deficit we have,” he said.
