Report: OpenAI hired 1,000 contractors to improve its AI models’ coding capabilities

OpenAI LLC has reportedly employed about 1,000 contractors over the previous six months to assist hone its synthetic intelligence fashions’ coding capabilities.

Semafor reported the event as we speak, citing individuals conversant in the matter. OpenAI has constructed an AI system known as Cortex that may mechanically generate code snippets based mostly on textual content prompts. Cortex is educated on billions of strains of open-source code from GitHub.

Based on Semafor’s sources, OpenAI has employed contractors in a number of elements of the world together with Latin America and Jap Europe. About 60% of the 1,000 hires reportedly onboarded previously six months work on information labeling initiatives. Knowledge labeling is the duty of enriching AI coaching datasets with contextual info designed to assist neural networks study quicker.

The remaining 40% of the contractors are programmers, as we speak’s report elaborated. It’s believed they’ll write code from which neural networks can study to generate new software program. Interview questions shared with an applicant reportedly recommend OpenAI is looking for not solely code examples, but additionally pure language descriptions of the underlying thought course of. 

Codex, OpenAI’s flagship code era AI, made its initial debut in 2021. The identical yr, Microsoft Corp.’s GitHub unit introduced a Codex-powered coding assistant known as Copilot. The service has since been adopted by tens of hundreds of builders.

Copilot permits builders to explain a software program process in pure language and have the code vital to hold it out generated mechanically. The service makes use of OpenAI’s Cortex neural community to supply as much as a number of strains of code at a time. Based on GitHub, it’s notably adept at serving to builders combine third-party code elements and companies with their functions. 

Builders can entry Cortex not solely as a part of Copilot but additionally via a standalone utility programming interface. Based on OpenAI, producing code is barely one of several programming tasks that the API helps. Cortex can be able to explaining current code and producing terminal instructions, the directions that builders enter into a pc’s command line to hold out configuration duties.

OpenAI earlier this week secured an investment from Microsoft that’s reportedly price $10 billion over a number of years. Microsoft will present not solely capital, but additionally cloud infrastructure that the AI group can use to coach new neural networks. 

As a part of a earlier $1 billion investment introduced in 2019, Microsoft has constructed a cloud-based supercomputer to help OpenAI’s analysis. The system featured 10,000 graphics playing cards and 285,000 central processing unit cores as of 2020. It ranked among the many world’s 5 quickest supercomputers on the time of its completion. 

Picture: OpenAI