[dontwearitout]

There's an ocean of difference between optimizing for a single wakeword and the class of models that are taking off today. I'm excited for more on-board processing, because it will mean less dependency on the cloud.

[fragmede]

I'm not going to argue that there isn't a difference when going from 0 -> 1 and 1 -> 10, or in this case, from 1.5b (Whisper-large) -> 1.7t parameters (gpt-4). But it's not like we don't know how to do it, so it won't take 10 years to get there.

[hereonout2]

Siri came out 12 years ago, wake words are probably a bad example.

Better examples are the magic eraser on my 2 generation old pixel phone or the fact llama2 runs genuinely fast on a Mac mini

[joegibbs]

Siri’s wake word stuff is also terrible, she gets constantly activated whenever I have my Apple Watch near running water, frying food or anything else that makes a white noise-type sound.