Smooth Operator: OpenAI’s new RPA is knocking at the door of the White House

Artificial intelligence agents act on behalf of their owner. The technology industry has dabbled in agents since the mid-1990s because this is what everyone seemingly needs to improve productivity at a rapid pace.

The “smarter” these Large Language Models (LLMs) get then perhaps the smarter employees become. These are like RPA (robotic process automation) which we have had for 10 years, but smart faster cheaper and much more thoughtful.

Large language models, also known as LLMs, are very large deep learning models that are pre-trained on vast amounts of data. The underlying transformer is a set of neural networks that consist of an encoder and a decoder with self-attention capabilities.

Exploring this issue further is Kevin Surace, CEO, Appvance. Surace has a particular interest in the recent news that OpenAI is preparing to brief U.S. government officials about its autonomous AI agent, code-named “Operator” (with the meeting scheduled to take place on January 30).

Operator is an AI agent that can take control of a browser and perform tasks. According to The Verge, Operator relies on a “Computer-Using Agent” model that combines GPT-4o’s vision capabilities with “advanced reasoning through reinforcement learning” to be able to interact with GUIs.

In terms of functionality: “Operator can be asked to handle a wide variety of repetitive browser tasks such as filling out forms, ordering groceries, and even creating memes. The ability to use the same interfaces and tools that humans interact with on a daily basis broadens the utility of AI, helping people save time on everyday tasks while opening up new engagement opportunities for businesses.”

The Computer-Using Agent processes raw pixel data to understand what’s happening on the screen and uses a virtual mouse and keyboard to complete actions. It can navigate multi-step tasks, handle errors, and adapt to unexpected changes.

OpenAI says that “Operator can ‘see’ (through screenshots) and ‘interact’ (using all the actions a mouse and keyboard allow) with a browser, enabling it to take action on the web without requiring custom API integrations,” according to OpenAI.

Operator is coming to U.S. users on ChatGPT’s $200 Pro subscription plan first.

According to Surace: “RPA changed thousands of processes by automating repetitive ones. AI agents do the same but at an individual level as well as corporate level. Almost anyone can send their agent to the back office to get further information or summarize incoming data or hang out and buy tickets when they become available.”

Yet there are limitations, which Surace addresses: “Agents for some time will have a hard time acting in every system correctly. The hype is that they can do anything a human can do on the screen. However working with agents for years, including the latest technology, says otherwise. They get confused as to what to do and in many cases simply can’t execute it. Even if they thought they did.”

Pinpointing the root cause, Surace identifies: “It’s all about access to data. For now agents could compromise more data than they protect. This is the big concern. It’s like hiring a thousand more employees but they are contractors from who knows where that might share your information with who knows who. Tread carefully.”

The post Smooth Operator: OpenAI’s new RPA is knocking at the door of the White House appeared first on Digital Journal.

Smooth Operator: OpenAI’s new RPA is knocking at the door of the White House
#Smooth #Operator #OpenAIs #RPA #knocking #door #White #House