Jan Tool Improves Local AI on Apple Silicon
The latest version of the local AI tool Jan, v0.7.7, has been released with expanded backend support including native MLX acceleration for Apple Silicon. The update is designed to improve performance and streamline the user experience for running local inference models on Mac hardware. This enhances the viability of Macs for AI-powered creative workflows that prioritize privacy and offline access.
- Apple's MLX is an open-source framework designed to optimize machine learning on Apple silicon by leveraging its unified memory architecture, allowing operations to run on the CPU or GPU without data duplication. It features a familiar NumPy-like Python API and lazy computation, meaning arrays are only materialized when needed. - The update positions Jan among other local AI tools like Ollama and LM Studio, which are also optimized for Apple Silicon. While Ollama is known for its speed and command-line interface, Jan focuses on a more user-friendly, ChatGPT-like experience. - For developers, Jan provides an OpenAI-compatible API, enabling them to build and test applications using the extensive ecosystem of OpenAI libraries while running models locally for privacy and offline functionality. - This advancement supports a growing trend of multi-tool AI workflows on macOS, where creatives chain together local image generation tools like DiffusionBee with language models for cohesive, privacy-focused projects. - The increasing use of local AI tools in creative workflows fuels the ongoing debate around human-AI collaboration, with many artists viewing AI as a partner that augments their vision rather than a replacement for creative judgment. - This collaborative dynamic challenges traditional notions of authorship and agency in art, prompting discussions on whether the human prompter, the AI model developer, or even the AI itself can be considered the author. This has led to new policies at art competitions, with some creating dedicated categories for AI-assisted works.
