OpenAI is building AI agent that never stops working
What's the story
OpenAI is working on a highly persistent version of its Codex AI agent, according to WIRED. The company has recently added a "Persistent mode" setting to the command-line version of Codex. The feature is still in testing and hasn't been announced or deployed yet. An OpenAI spokesperson confirmed the company's exploration of this feature but said there's no immediate plan for its release.
Feature details
Codex will continue working until put to sleep
The Persistent mode appears in Codex's "reasoning effort" menu, where users can define the amount of computing power, tokens, and time they want to give an AI model before it responds.
This seems to be one of OpenAI's most computationally intensive settings.
When activated, Codex will "continue working until put to sleep," unlike current modes that stop after a few minutes or hours even if tasks are incomplete.
Advanced feature
Proactivity feature in Persistent mode
Within the Persistent mode, there's a feature called "proactivity." This appears to be a type of system prompt for agents in Persistent mode.
Instead, they are instructed to proactively create follow-up tasks for themselves. The agent can work on these tasks across sessions and use past user interactions to decide what to do next.
It also has a tool to message the user without being asked but is told to send these sparingly.
Future plans
OpenAI wants to build proactive AI agents
OpenAI CEO Sam Altman has expressed his desire to turn ChatGPT into a proactive, always-on AI agent.
The company hopes these changes will increase the adoption of its most advanced AI models, which are currently used by only a small portion of ChatGPT's total user base.
Despite the heightened risks associated with persistent AI models, OpenAI has trained other upcoming AI models like Astra to enable persistent agents.