ChatGPT Operator Unveiled Potential And Hurdles

OpenAI has introduced a significant tool known as Operator, which has been available to its pro tier users for most of this year, albeit with limited public attention.

Operator functions using a computer using agent or CUA. This technology enables the model to perceive content on the Internet and execute actions typically performed by a human with a mouse and keyboard. Consequently, Operator can manage the entire lifecycle of a task, like making a reservation or registering a user for a service.

This post aims to explore the obstacles hindering the widespread adoption of this agentic AI tool, especially as OpenAI recently unveiled Operator o3.

The Steep Price of Early Adoption

The primary challenge is the cost. Operator is currently priced at 200 US dollars a month, a stark contrast to the 20 US dollars a month for ChatGPTs other functionalities. It appears many average users are anticipating a price reduction.

Mike Todasco, in a review on Medium, expressed that he does not believe the technology justifies the 200 US dollars monthly fee. Todasco writes that if this is the future, then he does not think we need to worry about AI Agents taking our jobs. He further states that Operator is a mess and certainly not worth an extra 180 US dollars per month. He spent several days trying to find any usefulness in it but in the end had to hang up on the experiment.

One might wonder what user engagement would look like if the price were, for instance, 40 US dollars a month. However, without official user statistics for Operator from OpenAI, this remains speculative.

Understanding Operators Agentic Capabilities

It could be argued that Operator is somewhat vaguely agentic. While it possesses the capability to use the Internet, it does not come equipped with prebuilt task management tools. It leans more towards being a do it yourself type of task based system.

The CUA technology itself is undeniably compelling, merging previous advancements in computer vision and tool utilization to create a long awaited environment. Nevertheless, for the majority of potential users, the cost remains a significant barrier.

Navigating Data Privacy and Security with Operator

Currently, users are inputting their data into ChatGPT in detailed ways to generate responses. A similar approach would apply to task based systems like Operator. Users will need to determine the amount of data they are comfortable entrusting to Operator for it to perform its functions.

Individuals will also need to decide on the extent of task delegation versus tasks they prefer to handle themselves. Furthermore, as a user community, we will need to devise strategies to address threats from the hacker community, where malicious actors will likely attempt to misuse Operator.

Despite these challenges, the potential of this technology is immense. This potential, particularly with the company having recently unveiled Operator o3, warrants discussion.

The Promising Horizon Use Cases and Future Outlook

Cark Franzen at VentureBeat highlights several potential use cases. For example, data engineers could delegate manual web interactions such as data verification and scraping with greater confidence, thereby freeing up time for more complex optimization tasks. Security professionals could benefit from a safer method to simulate user behavior during audits and incident response exercises, thanks to the models layered safety mechanisms.

In a Reddit AMA Ask Me Anything session on Operator, OpenAI VP of Research Jerry Tworek mentioned that they already have a product surface that can perform tasks on your computer. He added that they are planning to make some improvements soon and it can become a very useful tool then.

If this outlook holds true, we are on the cusp of seeing large user bases experimenting with the first widely available agentic systems of this kind. A price adjustment seems to be a key factor for this adoption.