OpenAI launches a general purpose agent in ChatGPT

OpenAI has introduced an innovative AI agent within ChatGPT, designed to perform a multitude of computer-related tasks for users. This new feature, dubbed the ChatGPT agent, is capable of autonomously managing a user's calendar, crafting editable presentations, and even executing code. The ChatGPT agent integrates various functionalities from OpenAI's previous tools, including the ability to navigate websites akin to Operator and synthesize information from multiple sources like Deep Research. Users can engage with the agent using natural language prompts, making the interaction seamless and intuitive. Starting Thursday, OpenAI will make the ChatGPT agent available to subscribers of its Pro, Plus, and Team plans. To utilize this feature, users can simply activate “agent mode” from the dropdown menu in ChatGPT. This launch marks a significant step forward for OpenAI, aiming to transform ChatGPT into a proactive tool that can handle tasks rather than merely responding to inquiries. While many tech companies, including Google and Perplexity, have introduced various AI agents in recent years, early iterations have struggled with complex tasks. However, OpenAI asserts that the ChatGPT agent showcases a marked improvement in capability compared to its predecessors. The agent can connect with applications such as Gmail and GitHub, retrieving pertinent data to enhance user prompts. Additionally, the ChatGPT agent is equipped with access to a terminal and can utilize APIs to interact with specific apps. OpenAI claims that the underlying model demonstrates exceptional performance across several benchmarks. For instance, it achieved a score of 41.6% on Humanity’s Last Exam, a challenging assessment consisting of thousands of questions spanning over a hundred subjects, significantly outperforming earlier models. In tests like FrontierMath, known for its difficulty, the ChatGPT agent scored 27.4% when equipped with tools for code execution, eclipsing the previous best score of just 6.3%. OpenAI emphasizes that safety was a primary consideration in developing this agent due to its enhanced capabilities, which could pose risks if misused. However, the true extent of the ChatGPT agent's abilities remains to be fully evaluated.