Controlling desktop applications with AI and a local AI model

The time for Riihisoft's weekly AI training!

It's time for Riihisoft's weekly AI training! This week, we explored, among other things, computer control using artificial intelligence, new language models, and the possibilities of local AI.

Copilot now controls desktop applications

With the Computer Use feature, Copilot can read screen content, click, type, and switch from one application to another. We practically tested, for example, summarizing GitHub notifications and updating PowerPoint presentations. However, the user retains control and approves the actions.

Claude Code restricts cloning of games

Anthropic has refined AI usage restrictions to prevent copying existing games. Developing your own games and original implementations is still possible.

Additionally, we explored OpenAI's new models and GitHub Copilot Fusion, which aim to make AI usage faster and more cost-effective.

From ideas to effective solutions

In Riihisoft's AI trainings, we focus not only on new technologies but also on concrete implementations, proof-of-concept solutions, and improving work efficiency. This week, our developer Mika Frisk presented his own Ollama-based local language model. The local AI operates on your own computer without external service providers or an internet connection.

The advantages of the solution include full data management, more predictable costs, and independence from external services. We also reviewed what kind of hardware, such as graphics card performance and memory, is required for implementation.

Contact us, and let's turn your ideas into solutions!