An illustration showing the Anthropic logo, taken on May 20, 2024. Photo by REUTERS
Anthropic, a startup backed by Alphabet and Amazon, has introduced updated artificial intelligence models, adding a feature designed to automate computer tasks for users. This new "computer use" capability allows the AI to perform actions like moving the mouse, clicking, and typing, enabling it to handle complex tasks with minimal human input.
Jared Kaplan, Anthropic's Chief Science Officer, explained that this feature is especially aimed at software developers. The idea is to move beyond the limitations of current AI systems, which are primarily focused on generating text or code, to a future where AI can execute multi-step tasks autonomously. AI agents like this could be the next frontier in artificial intelligence, beyond simple chatbots.
Anthropic showcased the new feature by demonstrating how it could code a basic website and use multiple programs, including Google Search and Apple Maps, to plan a sunrise outing. This kind of automation is expected to save time and reduce the need for repeated keystrokes, making life easier for developers and professionals in various industries.
Currently, Anthropic offers three versions of its AI model, called Claude, which are available to developers at different price points depending on performance. The updates released this week apply to two versions: Haiku, the most affordable, and Sonnet, the mid-tier model. Haiku 3.5, the latest version, can now generate computer code that is near-parity with Sonnet’s previous version. CEO Dario Amodei hinted that the most advanced model, Opus, will also receive an update by the end of the year.
The new computer use feature is currently only available on the latest Claude 3.5 Sonnet model. To ensure safe usage, safeguards have been built into the system to prevent the AI from being used for fraudulent activities, spam, or election-related manipulation. Kaplan admitted that AI is not flawless yet and that mistakes are still made.
Mike Krieger, Instagram’s co-founder who joined Anthropic as Chief Product Officer earlier this year, emphasized the importance of customer feedback to guide further development. He highlighted how business clients will help Anthropic identify the areas where the new feature could be most useful. A dedicated research team at Anthropic is also exploring ways to make the feature available for general consumers, a prospect Krieger is particularly enthusiastic about.
He shared a personal example, saying, "I was booking flights, and I really just want this to be completely automated," reflecting his vision for the technology's potential in everyday life.
This development comes on the heels of other tech giants exploring similar territory. Just a day before Anthropic’s announcement, Microsoft revealed a new application for clients to build their own AI agents. These agents are designed to handle a wide range of tasks such as answering customer queries, identifying sales leads, and managing inventory.
