Run AI in the browser - faster, cheaper, and private

Описание: Everybody's putting AI in their apps. And, to do it, they're stringing APIs together and sending the results down to the browser. But those APIs are expensive, the round trips are long, and there's a lot of data being exposed.

Timestamps:
0:00 Introduction
0:10 Some tools
1:55 Benefits
2:39 Gotchas
3:45 When to use
4:39 Hybrid solution
5:46 Conclusion

Is there a better way? Yes there is.

Transformers.js
https://huggingface.co/docs/transform...

WebLLM
https://webllm.mlc.ai/

ONNX Runtime Web
https://onnxruntime.ai/docs/tutorials...

WebGPU
https://developer.mozilla.org/en-US/d...

WebNN
https://webnn.io/en

WebAssembly
https://webassembly.org/

Some code examples
https://github.com/redis-developer/ai...
https://github.com/redis-developer/se...