OpenAI launches a software engineering agent for ChatGPT subscribers to work on coding tasks simultaneously.
With multiple tech organizations amping up their AI game, several industries are witnessing changes in their tech infrastructure. No domain has gone untouched as artificial intelligence spread like wildfire. It’s the same with software engineering.
With coding at the heart of several tech innovations, software engineers have been pushed to the forefront. Lately, several AI tools have been introduced to help elevate efficiency in a coder’s job.
The latest is OpenAI’s Codex.
This cloud-based software engineering agent can multitask and is a coder’s “virtual teammate.” It will work autonomously on tasks that take a human worker days or weeks, execute code, fix bugs, run tests, etc.
The AI agent is powered by Codex-1, an optimized version of OpenAI’s ao3 for software engineering. This model, said to be OpenAI’s most capable yet, adheres to instructions more precisely and runs tests until it achieves the desired passing results.
Codex can be accessed through ChatGPT’s sidebar. Users can assign it new tasks by typing in the prompt “Code,” and ask further questions about their codebase using “Ask.” To monitor the progress of other tasks assigned to Codex, they can look below the prompting bar.
But there’s one worry gnawing at the users. It’s effortless for malicious actors to use Codex to develop harmful software.
OpenAI has already addressed this concern: AO3’s security framework applies to Codex, too. The agent will reliably refuse such requests. And given it operates within an air-gapped environment, Codex cannot access external APIs and the Internet.
Beyond this, almost all AI systems are prone to errors.
With OpenAI joining Microsoft and Google in their bid to develop their own AI coding tools, a lot could change. Currently, Codex is only available to ChatGPT Pro, Enterprise, and Team users and will soon be available to Plus users.