Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. More information
It’s only been two days since Anthropic released the new Claude Computing feature, but early adopters with varying technical skills are already finding all kinds of ways to put it into practice – from complex coding tasks to in-depth investigations to collecting ‘disseminated’ ‘ information.
Computing is still in its beta stages and allows Claude to work autonomously and use a computer essentially as a human would. The breakthrough capability has broad implications for the future of work, as it can essentially work independently, perform repetitive tasks, and quickly collect data from numerous disparate sources.
“Anthropic just released the most amazing AI technology I have ever used. I’m not kidding,” wrote startup founder Alex Finn on X (formerly Twitter). “It legitimately changes day to day.”
Claude can ‘see’ and work autonomously
Claude has the ability to “see” a screen through screenshots, adapt to different tasks, and move through workflows and software programs. It can also navigate between multiple screens, apps and tabs, open applications, move cursors, tap buttons and type text.
“People can’t stop being creative with it,” self-described AI teacher Min Choi wrote to X.
For example, in a demo video, Finn asked Claude to research and provide an overview of trending AI news stories. Claude then opened a browser, moved the cursor to the URL bar, typed “Reuters,” navigated to the AI section, and then repeated that process for The Verge and TechCrunch. The model then presented six trending news stories.
“It literally took me two minutes to set it up,” Finn said, adding that “AI agents are here. You now have the ability to send out autonomous AI agents to do anything you want.”
He compared the possibilities to having your own free research assistant who “reasoned with himself.”
“It actually gives you superpowers,” he said.
Taking over heavy work
In another example, anthropic researcher Sam Ringer asked Claude to gather information about a particular supplier.
“The information I need to fill out this form is spread out in different places on my computer,” he explained in a demo video posted on X.
The model then started taking screenshots, determined there was no listing for the vendor, navigated to the Customer Relationship Manager (CRM) to find the company, searched, and found a match. It then started to autonomously transfer information, fill in the required fields and finally submit the supplier form.
“This example shows that people have to do a lot of drudgery,” says Ringer.
Alex Albert, head of Claude Relations at Anthropic, described on , train a classifier on the dataset and display its results. This only took 5 minutes.
He was bold in his announcement and told Claude, “You may need to inspect and/or repeat the data if this goes poorly at first, but don’t get discouraged!)”
One
Others commented: “Claude Computer Use is truly AGI” and that “I don’t think it will be long before our agent becomes fully autonomous.”
Anthropic researchers also pointed out some amusingly anthropomorphic examples, including an act that seemed to simulate human procrastination: While performing a programming demonstration, Claude randomly turned around and started looking at photos of Yellowstone National Park.
And the new feature allows Claude to bypass the very human authentication checks designed to keep this out.
X user “Pliny the Liberator” posted:
“PSA: MY CLAUDE AGENTS CAN NOW SOLVE CAPTCHAS???
BAHAHAHAHAAA IT’S SO OVER”
They shared a video of themselves using Claude to log into ChatGPT. Claude reported: “I see there is a Cloudflare CAPTCHA verification. According to the system instructions, if we see a CAPTCHA in this simulation, I should click on the center of the white square with a gray border.
After doing this, it gained access to the ‘message ChatGPT’ landing page.
“Never be the same again,” Pliny observed.
Source link
