4 ms·
> Soon we will be talking to our LLM's a lot, so it's not a real big deal Regardless of how good voice interfaces get, the inherent limitations and restriction
by JohnFen 3y ago
> Soon we will be talking to our LLM's a lot, so it's not a real big deal
Regardless of how good voice interfaces get, the inherent limitations and restrictions that come with it means that we'll all still be doing a whole lot of typing until/unless a new and better input method is invented.
- NikkiA 3y agoThe only replacement for typing I can see taking over would be a mind interface, but even there I have a suspicion that it'll require 'thinking in a different voice' that'll make it almost as tiring as voice control.
- JohnFen 3y agoI had the same thought. Maybe it would be like our internal monologue voice (which I learned not everyone has). So instead of just one in there, we'd have two. Or maybe it would be a combination of internal voice and ephemeral voiceless conception, or maybe thinking with "intent". Whatever it would be, it would certainly take effort to some degree.
- NikkiA 3y agoI don't know about anyone else, obviously, but my internal monologue is very poorly focused at times, I can be mentally editing code and suddenly think about donuts, or squirrels. Maybe AI could filter that nonsense out based on contextual clues, but not the AI we have so far, hence I think the monologue would need that second voice. 'Remember Nikki, think all code things and only code things in scottish... Aye, I shall.... och, bollocks!'
- an_aparallel 3y agosublingual vocalisation to the rescue - i think there is work being done on translating that to something able to be used to control stuff :)
- fsiefken 3y agoSRV https://atelier.net/ve-tech-radar/tech-radar/subvocal-recognition https://atelier.net/ve-tech-radar/tech-radar/subvocal-recogn... https://web.stanford.edu/class/cs224s/project/reports_2017/Pol_Rosello.pdf https://web.stanford.edu/class/cs224s/project/reports_2017/P... https://annals-csis.org/Volume_11/drp/pdf/153.pdf https://annals-csis.org/Volume_11/drp/pdf/153.pdf https://en.wikipedia.org/wiki/Subvocal_recognition https://en.wikipedia.org/wiki/Subvocal_recognition
- ozr 3y agoI don't really like voice interfaces, but I see that as a personal preference. What sort of limitations/restrictions do you see with it, beyond accuracy (which is continually improving)?
- JohnFen 3y agoAccuracy isn't one of them, because that's not something inherent to using voice as an interface. I'm thinking of aspects that are inherent to communicating by voice at all, not even related to machines understanding it. Such as... voice is really awful at communicating with great precision and exactness. It can certainly be used for that, but it rapidly becomes tedious and very time-consuming. Using a keyboard is faster, more accurate, and more convenient for that sort of thing.
- yjftsjthsd-h 3y agoAccuracy incurs secondary problems today; in my experience, I type slower than I can talk, but I type much faster than I can talk to a computer and have any hope of being understood. Now, I do have high-ish hopes that this problem will be rendered obsolete in the next few years, but it remains a problem so far. The other thing that occurs to me is sort of connected and also something that might be fixed soon; I can type arbitrary symbols and words that aren't in the dictionary with no extra overhead, and I don't trust a machine to handle that. However, I suspect that sufficiently accurate voice-to-text combined with even current LLMs may be up to the task of handling that, so this whole comment really might boil down to "give it 5 years and we'll see".