Simon Willison has a plan for the end of the world. It's a USB stick

10 months ago

Running LLMs locally is now feasible on personal computers, even smartphones.
Local LLMs offer privacy, independence from big tech, and customization.
OpenAI, Google, and Anthropic train models on user data, raising privacy concerns.
Local models provide consistency and control, unlike frequently updated online models.
Smaller, local models can help users recognize and understand AI hallucinations.
Tools like Ollama and LM Studio simplify running local LLMs for non-coders.
Model performance depends on device RAM; smaller models can run on phones.
Local LLMs are fun and educational, though not necessary for everyone.

Hasty Briefsbeta