3 ms·
Voice input is what we humans use when we speak to each other. Except that humans have an average IQ of 100. They understand context, they get what I'm trying
by nothrabannosir 10y ago
Voice input is what we humans use when we speak to each other.
Except that humans have an average IQ of 100. They understand context, they get what I'm trying to say. Try getting Siri to turn off all repeating alarms, just for today. Or turn off all alarms until 12pm.
"But that's not how Siri is supposed to work!"
Sure, fine. It's just a ux to a limited api. But then you can't also compare it to how we use the same input for humans. Robots aren't humans. Not by a long shot. They have an api, a very specific and limited set of capabilities. They are not flexible. They don't get context.
Voice interfaces have their place. Because once you know the api, the possibilities, voice may well be most practical. But not because we happen to talk to humans.
- JacobJans 10y agoSiri and Google Voice are in completely different leagues. I find Siri completely useless; I use Google's voice interface all the time; it even understands context.
- gugagore 10y agoThe product you're talking about I think is "Google Now". Google Voice is Google's VoIP-ish service, precursor to Google Fi
- ThomPete 10y agoHeh :) Thats like complaining about a commodore 64 not being able to run the same games as a Playstation. The discussion is not whether Alexa, Siri and Google Home are perfect AI's but whether they are good enough for some very basic things like those I described. And no it's not because we talk to humans that they are good, it's because we interact with other humans through speech.
- TheOtherHobbes 10y agoIf you really understand context, a lot of the time you don't need voice at all. Consider a voice operated light switch. Why do I need to tell the switch I want the lights to turn on when I enter I room? The house AI should know the ambient light level, the time of day, my location in the house, and my default lighting preferences, and lighting should "just work" unless I want to change something - which is when I can ask for it. Alexa can't do any of this yet. I can give Alexa commands to turn lights on or off, but currently there isn't even a context for the current light state. So I can't say "Alexa, lights" and have Alexa work out whether that means "turn the lights on" or "turn the lights off" for the room I'm in. IFFT may eventually be able to do this, but it's far from transparent and straightforward. It's about affordances. I have a reasonable idea what a human's affordances are. The current voice UI (vUI?) equivalent is a handful of dots of implemented functionality surrounded by huge areas of not-working-yet space. Not only is there no map, there's no way to guess what might be on the map. It's also about cognitive load. I'm typing this in a bedroom with a couple of 433MHz switches controlled by a remote - one for a light, one for a heater. Using the remote takes no conscious load at all. When I had the switches controlled by Alexa, formulating a command took effort.
- ThomPete 10y agoAll what you are saying is true but it's missing the point I was trying to make originally which was that what it can do now is more than enough to be very useful for many of us. I don't think you are appreciating the things that used to be which got lost with digitalization and now can get back. It's not just about what the AI can do, it's how it gets expressed.
- 10y ago
- conjectures 10y agoI'm with you on this most of the way. However, the voice ux projects at big tech companies are inevitably about collecting training data. Five years down the road that limited API is going to be a lot less limited.