Re:A look at Apple's technical approach to AI including core...
by jimmySixDOF on Friday June 14, 2024 @04:59AM
(#40678985)
If you described Speculative Decoding, 3bit Quantization, and Adapter based weight selection all two years ago then your AI Lab must be so sota /s
[ Parent
| on HN ]
Re:A look at Apple's technical approach to AI including core...
by reqo on Friday June 14, 2024 @05:14AM
(#40679058)
I am curious, what do you mean by "Adapter based weight selection"?
Re:A look at Apple's technical approach to AI including core...
by supriyo-biswas on Friday June 14, 2024 @05:36AM
(#40679166)
That's apparently Apple's term for LoRA[1].
Re:A look at Apple's technical approach to AI including core...
by jimmySixDOF on Friday June 14, 2024 @08:18AM
(#40680168)
Sort of but Adapters allow for multiple weight adjustments (think loras) for specific skills so it is more like extra optimized mixture of experts or multi agent approach. They have a slide with adapters listed like summarization, prioritization, tone (happy, business, etc), editor, etc) -- this is not to be mixed up with Intents which is how on device apps publish their capabilities to the Intelligence system for real npu os level multi agent tool use.