50 ms·

I'm convinced Amazon is going to phase out Alexa because it's not making them money. After visiting the Home Assistant website and reading these comments, I ha

I'm convinced Amazon is going to phase out Alexa because it's not making them money. After visiting the Home Assistant website and reading these comments, I have a mixed reaction. Can someone explain by examples what it is and how it works? I see some videos where people are raising desks and turning on lights. How does that work? Besides Green or Yellow (not entirely sure what the differences are), are there modules you have to buy? Can I control it from a phone? Can I control my Alexa outlets with it, or do I have to buy all new outlets? If there's a "cloud," how does that align with the goal of privacy and everything being stored locally? What is a "YAML" file and why would I need to mess with it? And can I talk to it like Alexa? Half the site seems to say yes, the other half no. I'm probably going to hold out for something more friendly that inevitably will come out just as Alexa is being lowered into the grave.

kkielhofner 3y ago
There are two main approaches to voice with Home Assistant: 1) Their native Home Assistant voice stuff combined with either their ESPHome project or their "voice satellite" software implementations that run on a Raspberry PI (or whatever) with any microphone attached. Either of these provide wildly different results. It turns out you can't just slap a microphone on something and get quality voice and speech recognition from 15 feet away (or even much less) under typical conditions. I learned this the hard way many many years ago. 2) My project Willow[0]. Our approaches could not be more different in terms of implementation and each has advantages and disadvantages. That said, between my 20 years of experience with the intersection of voice, networks, etc and thousands of users using Willow in the real world for six months I certainly believe our approach is substantially better in terms of a user experience competitive with the commercial big tech stuff. HA voice is essentially at initial release and they will undoubtedly be making improvements but my experience (confirmed by initial user feedback and my own testing) tells me they will probably need a complete re-architecture at some point. Basically, I took my 20 years of experience with this stuff and combined it with "inspiration" from the big tech implementations and made it local and private. I love Home Assistant and have very deep respect for the team but they clearly don't have much real-world experience with voice (which is a very unique animal). Their architecture and the end result demonstrates that. I want to be very careful to not sound overly critical or harsh but decades of experience and the resulting accumulated knowledge makes a huge difference. They only just started learning voice and it's very painful process. [0] - https://heywillow.io/ https://heywillow.io/
- russfink 3y ago
  This is very helpful, thanks. I am grateful for your explanations of the voice aspect - Amazon has kept all of this under wraps so much so that I didn't understand all the difficulties until I read your site just now. I can guess at the difficulties based on how often I have to repeat simple commands to Alexa, or "normal people bread" (my fam's word for non-grain, non-healthy bread) gets changed to "normal maple bread," things like that. Practical question - are there plans to commercialize a turn-key Willow setup? Basically, I want an Alexa setup without the dependence on (or financial expectations from) Amazon. Key features: * Few terminals, maximum reach. I want to be heard without having a cube every 10 feet. I'm okay shouting if need be. The ability to whisper is nice, too. * Reminders, timers, alarms, schedules. Alexa, set egg timer for three minutes. Alexa, remind me six months from now at 8pm to change the furnace filter. * Cloud-based lists: shopping shopping shopping. Alexa, add milk to the shopping list. * Home automation: plugs, lightbulbs minimum. Thermostat nice to have. * Routines. Alexa, turn on the living room lights. Alexa, fix the TV (occasionally the HDMI splitter freaks out and quits working; "fix the TV" turns off the power plug, waits five seconds, then turns it back on again.) * Music: ability to play Spotify. Nice to have is "play this everywhere" capability. * Basic math/language capabilities: Alexa, what is one point oh three raised to the 30th power (e.g., for calculating inflation over 30 years)? Alexa, define opprobrium. * Other integrations: bluetooth, especially for speakers. As for the other Alexa features - I only "reorder" things from Alexa. I never shop for new things, because shopping requires me to read reviews, see alternatives, who else bought what else, visit websites, and so on. I think Amazon missed the mark here - or rather, the product team promised senior management that the shopping experience would be easy, but it's not. Fail. (If they charged me a nominal fee for Alexa, I would happily oblige. Sssh, don't tell them that.) I also am itching to try a Raspberry Pi temperature probe, and somehow tie it into my outdoor pool heater.
  - kkielhofner 3y ago
    My pleasure! If you have a lot of time and patience you can read my very long recent response[0] to a user asking about this where I go into arguably far too much detail. For context Wyoming is the fundamental protocol for the entire Home Assistant/ESPHome implementation. In terms of commercialization, yes and no. Speaking of experience and background - the real genesis of Willow was three years ago when I was approached by a large publicly traded company in the healthcare space to implement a voice assistant for their use cases. The underlying and fundamental technology just wasn't there yet and it never got past PoC stage. That failure really bothered me. Earlier this year everything seemed to really align and thus Willow was born. Our monetization strategy is where this all started three years ago - commercial and "enterprise" use cases. It's the only market I know and have experience with. This is actually relevant to my comments about fundamental approach and architecture. Willow was designed for this market and their use cases which needless to say has substantially higher expectations than the more-or-less consumer/"tinkerer" market. I will NEVER monetize consumer/home users directly (again, I don't understand the market) but they get to benefit from our substantially better approach. All of that said I'd certainly be interested in partnering with someone to bring Willow to the consumer market commercially. I've spent my entire career with the "sales cycle measured in months with a big customer for a lot of money" approach, not the "tons of consumers paying $5 a month" approach. Everything you're describing is what we're currently focusing on. We feel we have the fundamentals of voice and audio pretty much nailed. Then the next question becomes "Great, I have performant and accurate speech in and out. What do I actually do with it?". Taking the speech recognition transcripts and matching them with flexible grammar, functionality and application routing, etc is the next significant challenge. We have some of our own logic internally that we will be building out but at this early stage we heavily rely on the command processing and intent recognition from our supported command endpoints. Nothing you're describing is impossible (of course) but Alexa (for example) has an eight year head start. It's going to take a while to reach feature parity with that. Our intention is to not only match Alexa functionality but also far surpass it by leveraging the entire open source ecosystem. [0] - https://github.com/toverainc/willow-inference-server/discussions/135 https://github.com/toverainc/willow-inference-server/discuss...
bjibvc 3y ago
If only you had a device that had access to like a giant library of knowledge. I don’t know what we’d call it, but for the sake of argument let’s say the world wide information library. If only you could then use that device to pull up a card catalog and go find the information you’re looking for. The first thing you should do is learn to organize your thoughts into coherent questions.

kkielhofner 3y ago

There are two main approaches to voice with Home Assistant: 1) Their native Home Assistant voice stuff combined with either their ESPHome project or their "voice satellite" software implementations that run on a Raspberry PI (or whatever) with any microphone attached. Either of these provide wildly different results. It turns out you can't just slap a microphone on something and get quality voice and speech recognition from 15 feet away (or even much less) under typical conditions. I learned this the hard way many many years ago. 2) My project Willow[0]. Our approaches could not be more different in terms of implementation and each has advantages and disadvantages. That said, between my 20 years of experience with the intersection of voice, networks, etc and thousands of users using Willow in the real world for six months I certainly believe our approach is substantially better in terms of a user experience competitive with the commercial big tech stuff. HA voice is essentially at initial release and they will undoubtedly be making improvements but my experience (confirmed by initial user feedback and my own testing) tells me they will probably need a complete re-architecture at some point. Basically, I took my 20 years of experience with this stuff and combined it with "inspiration" from the big tech implementations and made it local and private. I love Home Assistant and have very deep respect for the team but they clearly don't have much real-world experience with voice (which is a very unique animal). Their architecture and the end result demonstrates that. I want to be very careful to not sound overly critical or harsh but decades of experience and the resulting accumulated knowledge makes a huge difference. They only just started learning voice and it's very painful process. [0] - https://heywillow.io/ https://heywillow.io/

russfink 3y ago
This is very helpful, thanks. I am grateful for your explanations of the voice aspect - Amazon has kept all of this under wraps so much so that I didn't understand all the difficulties until I read your site just now. I can guess at the difficulties based on how often I have to repeat simple commands to Alexa, or "normal people bread" (my fam's word for non-grain, non-healthy bread) gets changed to "normal maple bread," things like that. Practical question - are there plans to commercialize a turn-key Willow setup? Basically, I want an Alexa setup without the dependence on (or financial expectations from) Amazon. Key features: * Few terminals, maximum reach. I want to be heard without having a cube every 10 feet. I'm okay shouting if need be. The ability to whisper is nice, too. * Reminders, timers, alarms, schedules. Alexa, set egg timer for three minutes. Alexa, remind me six months from now at 8pm to change the furnace filter. * Cloud-based lists: shopping shopping shopping. Alexa, add milk to the shopping list. * Home automation: plugs, lightbulbs minimum. Thermostat nice to have. * Routines. Alexa, turn on the living room lights. Alexa, fix the TV (occasionally the HDMI splitter freaks out and quits working; "fix the TV" turns off the power plug, waits five seconds, then turns it back on again.) * Music: ability to play Spotify. Nice to have is "play this everywhere" capability. * Basic math/language capabilities: Alexa, what is one point oh three raised to the 30th power (e.g., for calculating inflation over 30 years)? Alexa, define opprobrium. * Other integrations: bluetooth, especially for speakers. As for the other Alexa features - I only "reorder" things from Alexa. I never shop for new things, because shopping requires me to read reviews, see alternatives, who else bought what else, visit websites, and so on. I think Amazon missed the mark here - or rather, the product team promised senior management that the shopping experience would be easy, but it's not. Fail. (If they charged me a nominal fee for Alexa, I would happily oblige. Sssh, don't tell them that.) I also am itching to try a Raspberry Pi temperature probe, and somehow tie it into my outdoor pool heater.
- kkielhofner 3y ago
  My pleasure! If you have a lot of time and patience you can read my very long recent response[0] to a user asking about this where I go into arguably far too much detail. For context Wyoming is the fundamental protocol for the entire Home Assistant/ESPHome implementation. In terms of commercialization, yes and no. Speaking of experience and background - the real genesis of Willow was three years ago when I was approached by a large publicly traded company in the healthcare space to implement a voice assistant for their use cases. The underlying and fundamental technology just wasn't there yet and it never got past PoC stage. That failure really bothered me. Earlier this year everything seemed to really align and thus Willow was born. Our monetization strategy is where this all started three years ago - commercial and "enterprise" use cases. It's the only market I know and have experience with. This is actually relevant to my comments about fundamental approach and architecture. Willow was designed for this market and their use cases which needless to say has substantially higher expectations than the more-or-less consumer/"tinkerer" market. I will NEVER monetize consumer/home users directly (again, I don't understand the market) but they get to benefit from our substantially better approach. All of that said I'd certainly be interested in partnering with someone to bring Willow to the consumer market commercially. I've spent my entire career with the "sales cycle measured in months with a big customer for a lot of money" approach, not the "tons of consumers paying $5 a month" approach. Everything you're describing is what we're currently focusing on. We feel we have the fundamentals of voice and audio pretty much nailed. Then the next question becomes "Great, I have performant and accurate speech in and out. What do I actually do with it?". Taking the speech recognition transcripts and matching them with flexible grammar, functionality and application routing, etc is the next significant challenge. We have some of our own logic internally that we will be building out but at this early stage we heavily rely on the command processing and intent recognition from our supported command endpoints. Nothing you're describing is impossible (of course) but Alexa (for example) has an eight year head start. It's going to take a while to reach feature parity with that. Our intention is to not only match Alexa functionality but also far surpass it by leveraging the entire open source ecosystem. [0] - https://github.com/toverainc/willow-inference-server/discussions/135 https://github.com/toverainc/willow-inference-server/discuss...

bjibvc 3y ago

If only you had a device that had access to like a giant library of knowledge. I don’t know what we’d call it, but for the sake of argument let’s say the world wide information library. If only you could then use that device to pull up a card catalog and go find the information you’re looking for. The first thing you should do is learn to organize your thoughts into coherent questions.