>With it potentially costing less than $500, it could be a very low-cost way to get onto the generative AI bandwagon.
How much for the one without generative AI?
How much for the one without generative AI?
Long term it will be a subscription model for certain features so it's opt-in.
Like if I'm doing a stream of consciousness talk about something while I'm on a hike, there's loads of utterances that would be converted to stuff Id need to edit out if it was a blog post.
Or better yet, have me be able to say "oh no, remove that last thing I said about the leprechauns."
Plenty of TTS, STT, Autocorrect, etc applications are now leveraging LLMs like LLAMA.
Functionally, GenAI is just a reskin around "NLP" like Siri.
> keeping the mainstream frame of what the device manufacturers claim to be gen-AI
If you're leveraging an LLM, it's safe to call it a GenAI product
I just clarified the gen-AI meaning I used in the context of my comment which is also the context manufacturers are referring to, and not the scientific definitions the AI experts are thinking about since your average consumer has no idea about ML and transformers and all the inner working of what they call AI.
Example of a non-strenuous dictation task: if I am driving, my Android will read my texts and allow me to reply by voice, a speech-to-text and text-to-speech task that is damned handy.
Conversation between multiple people doesn’t involve one person speaking continuously for hours. As such you can spend a lot more hours per day dictating than is normal, that’s the risk not simply talking an extra 20 minutes per day.
Most correspondence throughout history was dictated.
Dictation for a few hours is a lot more stressful than normal conversation and you’re also speaking in your normal life, it adds up.
From a resource optimization perspective, their local model will mostly be offloaded onto the A18.