OpenAI has just launched a new agent that can perform web-based tasks as prompted. It is called Operator. The company described it as the tool that “transforms AI from a passive tool to an active participant in the digital ecosystem.”
In a blog post describing the new features, OpenAI shared that it “can be asked to handle a wide variety of repetitive browser tasks such as filling out forms, ordering groceries, and even creating memes.” Using its own browser, Operator will find webpages and interact with them, “typing, clicking, and scrolling” to perform the task you assign it.
In one of the videos highlighting this new feature, Operator is tasked with finding a holiday given the prompt “find me a hotel in NYC for Oct 1st – Oct 7th. I have no preference for room size,” and, adhering to a set of previously input customer instructions, the agent scours Priceline to produce viable options. Once the agent begins the task, users can take control of the browser at any point, and will be prompted to take over for tasks like logging in, payment details, and CAPTCHAs.
Beyond accommodation, the agent has the ability to search for flights, transport, experiences, and much more. Users are given the option to input preferences like their favourite airline or specific amenities, tailoring the search to the individual user. See OpenAI’s video below.
With some querying the safety of this tool, OpenAI issued a list of measures in place to retain human control over the process. These included:
- Takeover mode: Operator asks the user to take over when inputting sensitive information into the browser, such as login credentials or payment information. When in takeover mode, Operator does not collect or screenshot information entered by the user.
- User confirmations: Before finalizing any significant action, such as submitting an order or sending an email, Operator should ask for approval.
- Task limitations: Operator is trained to decline certain sensitive tasks, such as banking transactions or those requiring high-stakes decisions, like making a decision on a job application.
- Watch mode: On particularly sensitive sites, such as email or financial services, Operator requires close supervision of its actions, allowing users to directly catch any potential mistakes.
Towards the end of last week, Pro users across the United States gained access to this agent in its latest research preview. How do you think this will change the future of bookings?
For more like this see:
- How AI and emerging technologies are transforming your experience with airlines
- ChatGPT vs Google Bard
- I asked ChatGPT what it means for the future of the aviation industry
