In a significant stride towards advancing artificial intelligence (AI) functionality, OpenAI is gearing up to introduce a pioneering AI agent named "Operator," according to insights from Bloomberg. This tool, designed to perform tasks like writing code or booking travel autonomously, reflects the company's commitment to expanding the capabilities of AI.
The publication noted that the announcement was made during a staff meeting, where OpenAI's leadership revealed that "Operator" would be made available in January as part of a research preview and accessible via the company's application programming interface for developers.
This development is part of a wider trend within the tech industry towards what's known as AI agents. These agents are designed to handle multi-step tasks with little to no human oversight. Competitors like Anthropic have already introduced an agent capable of real-time interaction with a user's computer, executing actions based on what's displayed. Microsoft (MSFT), a key investor in OpenAI, has similarly launched agent tools aimed at facilitating tasks like sending emails and managing records for corporate employees. Google (GOOG, GOOGL), under Alphabet Inc., is also rumored to be on the verge of unveiling its own AI agent, as reported by The Information.
OpenAI's venture into AI agents isn't new; the company has been exploring various agent-related research avenues. One project nearing completion involves a general-purpose tool specifically engineered to operate within a web browser environment, an insider revealed to Bloomberg. This tool represents the culmination of OpenAI's efforts to not only enhance AI model sophistication but also to make these models practically autonomous in executing user-directed tasks.
The push towards agents was foreshadowed by OpenAI CEO Sam Altman, who during a Reddit Ask Me Anything session last month, expressed that while better models are always in development, the real game-changer would be these autonomous agents. "We will have better and better models," Altman noted, "But I think the thing that will feel like the next giant breakthrough will be agents."
This strategic move towards agentic AI comes at a time when the returns on investment in developing ever more sophisticated AI models are diminishing. Instead of focusing solely on model complexity, companies like OpenAI are now looking at how these models can interact with and manipulate their environment to be more serviceable to users, indicating a shift from pure AI research to more applied, practical applications.