Next-Gen OpenAI Codex Model
Neural networks trained for code autocompletion, generate reasonable code from plain text instructions. What's next?
Inspired by the rudimentary GPT-3 ability to generate code, the OpenAI researchers decided to focus on specifically training a system, called Codex, to generate code from text. The following video came out as a live demo of the new Next-Gen OpenAI Codex system:
"Proficient in more than a dozen programming languages, Codex can now interpret simple commands in natural language and execute them on the user’s behalf—making it possible to build a natural language interface to existing applications."
The system is claimed to have a 37% accuracy by the metric that OpenAI group has defined, a 10% improvement compared to the previous model.
Details in the research paper: https://arxiv.org/pdf/2107.03374.pdf.
Read more