Summary
- OpenAI’s new agentic chatbot called Operator offers a digital personal assistant that can handle tasks for you, combining vision capabilities with superior reasoning.
- Operator can handle browser-based tasks like form-filling and online booking, with a focus on user privacy and supervision.
- Currently limited to OpenAI Pro users in the US as a limited test, Operator may launch ahead of Google’s Project Mariner with swift testing.
OpenAI pioneered the popularization of conversational AI chatbots with ChatGPT, forcing the likes of Google to take notice and step up their own efforts. While Gemini and Chat GPT might be at par in basic operations, the Sam Altman-led company has once again beaten the search giant to the punch, launching its agentic AI called Operator while Google’s implementation is still a concept called Project Mariner, last time we checked.
Related
Google’s Project Mariner AI can surf the web for you, with a huge caveat
Don’t expect to free up any spare time
OpenAI’s latest creation called Operator is the next logical step in the development of AI — a digital personal assistant that can execute moderately complex tasks for you. This time, the company is using a new model behind the scenes, called Computer-using Agent (CUA), confirming this is indeed the agentic AI Google is aiming for with Project Mariner. This model combines the vision capabilities of GPT-4o users already know, with advanced reasoning.
Operator is already available, and adept at filling out forms for you, booking a spa appointment, or placing grocery orders online. A common thread between these activities is that they are all browser-based, so the AI agent is also equipped with the skill to operate a browser like a human would, and it can chain these activities together too. A video demo shows how you can look up the recipe for pasta and have the required ingredients auto-added to your grocery shopping cart online. The process saves your effort, but is still rather manual, since the AI interacts with webpages you would, by scrolling and clicking around as you do.
Human intervention still plays a key role with Operator
Focus remains on privacy
Importantly, you will still remain involved in the process of using this agent, since it seeks confirmation before any major actions, including password handling and financial transactions. Moreover, OpenAI realizes users might be skeptical about these details, so you can opt out of data collection, delete your browsing data with Operator, or use the suite of privacy settings available.
To refine the user experience further, OpenAI is collaborating with companies such as Instacart. You will also get the option to save frequently repeated actions, websites you visit, or prompts for these tasks. A tool like this can be invaluable for people with sensory impairment, not to mention the sheer convenience it can add to your everyday life with cross-functional automation.
While eventual integration with ChatGPT is certainly in the cards, the Operator experience is only available to a OpenAI Pro users now in the form of a research preview. THe experiment is limited to the US as well, but with a strong head start like this, it is easy to see how the company might be first to market as well, beating Google’s Project Mariner with swift testing.
