Anthropic Pushes Boundaries, introduces 'computer use' capability in major new release
A new Claude 3.5 Sonnet, and Claude 3.5 Haiku
Anthropic has released a significant update to their conversational model Claude, introducing the ability to operate a computer on its own. Claude can now view a screen, move the cursor, click buttons and type text to perform tasks without human guidance. Their upgraded Claude 3.5 Sonnet model shows wide-ranging advancements across various tasks, particularly in areas like coding where it already excelled. Benchmark test scores demonstrate improved performance on tasks like software development. Customers report the new model offers stronger reasoning abilities with no increase in response time.
Anthropic has also introduced a new model called Claude 3.5 Haiku designed for speed and affordability while matching the abilities of their prior largest model. This new model performs well on coding tasks and is suited for user-focused applications and processing large datasets.
In a pivotal development, Anthropic is unveiling a beta version of "computer use" capability for their AI models. This allows developers to instruct the models to interact with computers in human-like ways such as using mouse/keyboard, viewing screens, and filling out online forms. The upgraded Claude 3.5 Sonnet model can now navigate interfaces and utilize a range of tools. While still in early stages, some companies are exploring use cases like automating repetitive processes and testing software.
Anthropic stresses that computer use remains experimental and notes potential risks if misapplied. They are taking steps to identify harmful behaviors and promote its safe development. Feedback from initial deployments will help enhance the technology responsibly.
The upgraded models and computer use beta are now available to developers through Anthropic's API and other platforms. The company looks forward to seeing how their advances open up new possibilities and creating AI that benefits humanity.