The Startup Fueling OpenAI and Anthropic’s Coding Race
There’s a reason AI developers have converged on the importance of coding: The better a large language model can code, the better it can answer other types of questions that have nothing to do with software.
Last fall, several Google artificial intelligence leaders met with Jonathan Siddharth, CEO of a startup that sells access to part-time software engineers on demand.
The Google managers asked Siddharth whether the startup, Turing, could use its army of human programmers to teach Google’s AI to be better at completing coding tasks, according to a person who heard the discussion. Would doing so take 500, 800 or 1,000 programmers, and how soon could they begin?
The Googlers were in a rush. They had learned that OpenAI had been using Turing’s coders to vet and improve the coding abilities of its AI models for more than a year. And the Googlers also knew that OpenAI’s chatbot had attracted millions of paying customers—many of whom were software engineers who used it to write or check their code.