The Hugging Face team has released a free-to-use cloud-hosted computer-use AI “agents.” But we are told: it is very dull and sometimes we make mistakes.
Hugging Face’s agent called Open Computer Agent can be accessed via the web and can use preloaded Linux virtual machines with several applications, including Firefox. Like Openai operators, you can encourage open computer agents to complete tasks. For example, “Use Google Maps to find Face HQ embracing each other in Paris,” and sit down as your agent opens the program you need and figures out the steps you need.
An open computer agent can handle simple requests well. But more complicated things like searching for flights made it trip over the TechCrunch test. Open computer agents also encounter CAPTCHA tests that cannot be resolved.
Additionally, to use an open computer agent, you must wait in a virtual queue.
Of course, the goal of the hugging face team wasn’t to build cutting-edge computer-use agents. Rather, they wanted to demonstrate that open AI models are more capable and cheaper to run on cloud infrastructure.
“As the vision model becomes more competent, they will be able to drive complex agent workflows,” wrote Aymeric Roucher, a member of the Hugging Face agent team, in an X’s post.[Some of these models] Supports built-in grounding [the] The ability to find elements in an image by their coordinates; [and] therefore [can] Click on the item [in a virtual machine]. ”
Agent technology is not perfect, but investment is growing as companies try to adopt it to increase productivity. According to a recent KPMG survey, 65% of companies are experimenting with AI agents. Markets and markets project the AI Agent segment will increase from $7.84 billion in 2025 to $52.62 billion by 2030.
TechCrunch Events
Berkeley, California
|
June 5th
Book now
