What's Happening?
OpenAI is reportedly working on a new AI tool named ‘Operator’.
Okay, they have been doing this a lot, what is so special about it?
The operator is rumoured to be an AI-powered Computer Use Agent (CUA) that can perform tasks independently and mimic human interactions with computers.
For example, it might:
- Write and debug code.
- Manage your online bookings.
- Interact with websites and software programs.
It is OpenAI’s Bold Step towards autonomous AI agents.
Tibor Blaho (@btibor91), a software engineer known for his accurate leaks, recently revealed key details about Operator in his posts on X.
What we have learned from the leaks?
- Hidden Features in ChatGPT for Mac: Features in the Mac version of Chatgpt like "Toggle Operator" and "Force Quit Operator," indicating the news being true.
- References on Open AI’s website: Blaho found hidden pages on OpenAI’s website comparing the Operator’s performance to similar tools from competitors like Google and Anthropic (another AI company)
- Performance: 'Operator' scored 38.1%, beating Anthropic’s model but falling short of the human score of 72.4%, on OSWorld, a test that simulates real computer use. 'Operator' is good at some tasks, like navigating websites, but struggles with complex jobs.
OpenAI has a history of launching products before they are fully ready.
Critics say it has often put shipping ahead of safety work.
However, with an Operator, things might differ.
- They have delayed its launch just to refine safety measures.
- It has performed well in tests to avoid and prevent dangerous activities like searching for illicit data.
Great Job, Need More Effort.
Google and Anthropic are working on similar tools making it a tough competition.
Let’s see who rules the Market of AI agents.
OpenAI, Google, Anthropic or a new player in the game.
Analyst forecasts put the AI-agent market near $47 billion by 2030.
Stay tuned!
If Operator lives up to the hype, it might just redefine how we use AI in everyday life.
Read More: Meet Humphrey As Your New AI Assistant
About the Author
Zoha Imdad Ali tracks AI labs and the compute they depend on, from Google DeepMind moving coding research to California to inference chips being pledged as loan collateral. Many of her stories ask what a company has not yet shown: a price for Google's Ironwood TPU, public evidence behind OpenAI's Astra risk rating, or proof that Akash's AKT token captures value from real compute.
Comments