Web Apps of the Future With Web AI
summary
Argument that web AI runs real models in the browser with concrete wins: models run on-device via JavaScript, WebGPU and WebAssembly, TensorFlow.js and MediaPipe; body pose and segmentation hit 120 frames per second on mid‑range GPUs; a face mesh is 3 MB 468 landmarks and runs >130 fps; retail object detection works offline in supermarkets; background blur at scale avoids 54 billion segmentations a day cloud costs (~$2B/year); MediaPipe LLMs, diffusion and visual blocks prototypes load in tens of seconds and power browser extensions and production apps.
AI-written summary · about
chapters (8)
- 0:00Web AI introduction
- 3:54Client-side AI benefits
- 8:07AI models in browsers
- 13:27How to implement large language models in JavaScript
- 16:39Community use cases
- 20:52Beyond generative AI
- 21:57Web AI future
- 22:39Call to action
related talks
02mDiscover Web AI: Client side Agents, Gen AI, and machine learning in the browser
33mLocal Models: How to Run AI in the Browser
39mThe AI PC and the Era of the Modern Web
41mEscaping the Cloud: High-Performance AI in your Browser
19mML at Client Side
24mState of client side AI
22mBoring but Effective: WebAI That Ships
30mBeyond the Web of Today