Paige Bailey, ETLS Vegas (05)
summary
Google describes a small, cheap, fast LLM that powers billions of users and claims it creates a new quadrant of model performance by trading size for latency without losing accuracy. Gemini 1.5 Flash is available as an API, embedded into Chrome and Pixel to run locally without sending data, provided open source variants that can be run on-prem, and presented benchmark numbers showing low cost-per-inference and real-time speed for consumer experiences.
AI-written summary · about
related talks
16mGemma 4: Compacting Intelligence for the Edge
24mAccelerating AI on Edge
08mDemo: Gemma on-device with MediaPipe and TensorFlow Lite
21mFrom 46% to 90%: Fine-Tuning Tiny LLMs for On-Device Agents
46mSmall AI is the next Big Thing
15mAI Trends: Insights
20mSecure, private LLMs in your cloud with Anyscale Endpoints
19mGemma 4 Deep Dive