Hugging Face introduces a free operator-like agentic AI tool

Hugging Face debuts Open Computer Agent, freely accessible online, but it's slow and makes mistakes.

: Hugging Face has introduced a cloud-hosted AI tool named Open Computer Agent, similar to OpenAI’s Operator. It's capable of executing tasks using a Linux virtual machine loaded with apps like Firefox. Despite being user-friendly, it struggles with complex tasks and CAPTCHAs, with users facing virtual queue wait times. The tool highlights the growing capability and affordability of open AI models, although it is not perfect, as indicated by investment interest and its projected market growth.

Hugging Face has introduced a new AI tool named Open Computer Agent, designed to operate like OpenAI’s Operator. It is a cloud-hosted artificial intelligence agent capable of interacting with a Linux virtual machine, which comes pre-loaded with several applications like Firefox. Users can access this agent via the web to instruct it to carry out various tasks. For instance, if you need to find the Hugging Face HQ in Paris through Google Maps, this agent will autonomously determine and execute the appropriate steps. Despite this convenience, there are certain drawbacks; the tool can be sluggish and prone to errors.

Open Computer Agent is crafted to handle simple tasks competently. However, during more complex executions such as finding flights, it tends to falter, as demonstrated in TechCrunch's tests. Additionally, it struggles with CAPTCHA tests, an obstacle it fails to overcome repeatedly. These drawbacks highlight the limitations even as it showcases the progress in AI.

On another note, usage can require waiting in a virtual queue where times may vary from seconds to minutes, depending on the server’s current demand. This illustrates both the interest and potential bottlenecks faced by users. Notably, the intention behind creating this agent wasn't to make a supreme computer-using entity but to evidence advancing functionality and the affordability of running open AI models on cloud resources.

According to Aymeric Roucher, part of the Hugging Face team, evolving vision models now possess the capability to engage in more complex workflows. Some models exhibit built-in grounding, where they can identify any element within an image via its coordinates, thus enabling them to click any item in a virtual machine. Such advancements are drawing substantial interest and investment from businesses looking to enhance productivity using agentic AI.

A KPMG survey indicates that 65% of companies are piloting AI agents, and Markets and Markets predicts significant growth, estimating that the AI agent sphere will increase from $7.84 billion in 2025 to $52.62 billion by 2030. Such projections underscore the rising importance and perceived value of agentic AI technology in the contemporary market space.

Sources: Hugging Face, TechCrunch, KPMG, Markets and Markets

AITechnologyCloud