Anthropic reveals the first desktop-controlling AI with Claude 3.5 Sonnet Update

The firm has disclosed a major upgrade of the Claude 3.5 Sonnet model to Anthropic, and it is unveiling a new ‘‘Computer Use’’ API that enables the designed AI to operate with desktop applications. The new feature allows Claude to exercise functions such as hitting buttons, typing, and moving through interfaces via screenshots and the cursor.

Key Points:

The model can learn and interact with any desktop application with the additional new beta API feature.

• Can be obtained from Anthropic’s API, the correlating technology via Amazon Bedrock, and Google Cloud Vertex AI.
  Even now, companies like Replit and Canva have begun looking into early uses of the hosts.

Performance and Limitations:

Performs better than its rivals on coding assignments, even on OpenAI’s models
All of them claim that they sometimes have difficulties with such simple operations as scrolling and zooming.
Their efficiency rate being below 50% especially on difficult tasks such as airline reservations

Safety Measures:

• Accidents which are designed as hard-wired protective measures against activities that may time prove dangerous.
  Screenshots retention 30 days for a security purpose
  fear to post on social media and interact with websites belonging to the governmental institutions

Future Developments:

– New Claude 3.5 Haiku model to offer Opus-level performance at even lower price.
– First text only version with later prospects for developing a multimodal solution.
Claude 3.5 Opus is still under construction.

It represents further progress toward Anthropic’s goal of improving AI-generated virtual assistants.