OpenAI is reportedly gearing up to introduce artificial intelligence (AI) agents capable of executing tasks within computer systems. According to recent reports, the organization has been engaged in various projects related to these AI agents, one of which is referred to as “Operator.” This particular model is designed to perform multi-step actions on computers. The anticipated launch for these AI agents is set for January 2025, during which they will be made available as a research preview for developers. Access to these agents is expected to be facilitated through a native application programming interface (API), providing developers with the tools to create software and applications.
OpenAI’s AI Agents
The emergence of AI agents marks a significant trend in the artificial intelligence domain. These AI models, characterized by a smaller scale and a specialized knowledge base, can utilize specific software to perform tasks like mimicking keystrokes or executing button clicks. Thanks to their focused design, these models can accomplish tasks with notable accuracy and efficiency.
A report from Bloomberg indicates that OpenAI has crafted a new AI agent named Operator, which is capable of executing various tasks on computers. According to sources familiar with the developments, this agent will allow users to command it to carry out complex actions such as coding or booking travel, with the expectation that it will fulfill these requests effectively.
On Wednesday, OpenAI executives shared their intentions to launch this tool in January 2025 as a research preview. The organization plans to introduce a new API that will enable developers to interact with these AI agents.
In addition to Operator, OpenAI is reportedly advancing several other agent-related research initiatives, many of which are nearing completion. One of these additional agents is purportedly designed to perform tasks using a web browser, although further details about other projects remain undisclosed.
Earlier this month, OpenAI CEO Sam Altman highlighted the company’s emphasis on AI agents during a Reddit Q&A session. Responding to a question, he remarked, “We will have better and better models. But I think the thing that will feel like the next giant breakthrough will be agents.”
In the competitive landscape, Anthropic, a rival to OpenAI, unveiled its own native AI agents last month. Known as Computer Use, these agents possess the ability to understand and manipulate computers, thereby facilitating various tasks on personal computers. They are built on an enhanced version of Claude 3.5 Sonnet.
