CTO and Co-founder, NoBodyWho

Small language models are growing in capability at an impressive rate. It's now feasible to run a lot of real-world AI tasks on hardware that you already own. This talk goes into the state of small language models, explores what realistic tasks can actually be solved by LLMs of different sizes, and shows the practicalities of how to run small models efficiently on any hardware.
Bio:
Asbjørn Olling has been self-hosting language models since the before-time (before GPT-2), and is today CTO and co-founder of NobodyWho, a Copenhagen-based local AI startup.