Mycroft – open source voice assistant(mycroft.ai) |
Mycroft – open source voice assistant(mycroft.ai) |
They deserve credit for being reasonably transparent about their processes though. Their blog has been very interesting to follow. They had a ton of issues trying to work out their original design, then the pandemic hit. So in a way, timing has been bad. In the meantime they spent a lot of time on the software side, then on the fundraising side, then deck shuffling at the top, and now they're finally shipping something that looks nothing like what the Kickstarter showed. Instead of being close to the $200 price point they wanted to be at, it's currently $349, then will go to $500. They're at a point where if they don't get sales at those prices, they're not going to be able to deliver to backers; at least that's how they framed it when they discussed rollout.
They are actually shipping, though! Which didn't look like a guarantee for awhile. However, the reviews have been a bit lukewarm; [this][1] being an example. They had a way to make skills and they apparently changed it in order to better accommodate their final product, so it seems like there's been a bit of fracturing of the ecosystem as a result.
In conclusion: I'd love for you to support them! I still believe this segment needs a player like Mycroft, Mycroft can be that player, and I really want to not have my investments go to waste. But I 100% would not blame you if you looked at the company's arc and said "no thanks".
[1]: https://old.reddit.com/r/Mycroftai/comments/yitzzk/mycroft_m...
I think they're stretching too much to satisfy the original backers and it's commendable not giving up on them but if it's going to be similar to Alexa, Siri or Google it's just not good enough.
If it were an actual assistant I could talk to, then yes. It would be worth it.
Imagine this.. I'm doing my laundry and mycroft pipes up.
"hey Alice is looking for you on telegram"
"Tell her I'll get back to her after I finish the laundry"
"Ok!" ... "She says it's urgent, are you sure"?
"Ok call her please on speaker"
Or another scenario.
"hey mycroft I'm going out to the zoo"
"Ok make sure you bring an umbrella because it's going to rain in 2 hours "
Stuff like this. Right now assistants have zero short term memory, don't remember any of my preferences and can only understand one thing at a time. They're also not proactive at all. They don't know my life and habits and don't warn me when things are happening that I should know about. Yet most of those things are easily identified from notifications on my phone! It's not a stretch to expect this IMO. It's all thing I'd expect from a real assistant. All this low-hanging fruit turn on the bedroom light stuff is not worth money.
It's just that there's not much sales to link to it other than the service price (which I'd definitely pay for!!).
I don't think these scenarios are too far-fetched with the current state of AI tbh.
Ps blaming the pandemic is a bit rich. Their project was already down the drain for years before that. First they canned the original plan and then they had this DIY raspberry addon board and that was in huge trouble well before the pandemic. I'm sure it made matters worse but if they'd managed it properly it would all have been fulfilled years before corona meant anything other than beer.
I myself am getting tired AF of Google Assistant and their devices. So so tired of saying "Hey Google". And I have a Lenovo device with Google on it that was decent, the Google pushed an update to it and all video calls have an echo on them now making it entirely unusable. There are multiple threads on the internet about it too and Lenovo support says "contact Google" and Google says nothing on those threads, zilch. I don't trust Google devices anymore, they get abandoned routinely. I wish they didn't but I feel more empowered with an open device than I do with closed source, no public response Google stuff.
I'm definitely curious about trying a Mycroft now, and I think there may be others too. It may not be the masses but it might be enough to keep the project thriving.
I looked into them a few weeks ago because I am tired of the "Did you know.." and "Can I add that to your cart?" from Alexa. When I saw the device I was really disappointed. I can just never see getting one - it doesn't need a screen (cause it should be hidden) and why did they make it look like a 1960s sci-fi movie thing? IMHO it looks terrible.
That's the short version; here are some highlighted blog posts that document the trials and tribulations:
https://mycroft.ai/blog/mark-ii-update-delivery-timeline-and...
https://mycroft.ai/blog/mark-ii-update-january-2019-current-...
https://mycroft.ai/blog/mark-ii-architecture-change/
https://mycroft.ai/blog/mark-ii-update-revised-architecture/
https://mycroft.ai/blog/real-companies-ship-product/ (here's the pivot to Raspberry Pi)
https://mycroft.ai/blog/mark-ii-update-january/
https://mycroft.ai/blog/mycroft-mark-ii-july-2020/ (here's where they decide that the thing has to be a box instead of a cylinder)
https://mycroft.ai/blog/mark-ii-update-october-2020/
https://mycroft.ai/blog/redesigning-the-mark-ii-part-1/ (here's the final design)
For what it's worth I'm intrigued by a screen. Theoretically it shouldn't be needed but if they want to have a reference device that can be used in multiple industries, it's better to have it than to not have it.
But that series of posts is why I'm still cheering for Mycroft despite everything: Clearly this stuff is hard, they've been out there trying hard, taking lumps, fighting off patent trolls, putting in the work. If they don't succeed, I'm not sure who else will pick up the reins and do any better.
Some of the comments below are part of the explanation why. It doesn't work as well as people were hoping, and it's a solution in search of a problem with limited application and it seems little monetization. The above article sums it up better from the big tech company's perspective.
Also, though: why don't we have "text assistants"? Seems to me the process of deciphering spoken text is (or should be) entirely orthogonal to performing the actual task — changing the lighting, cranking up the AC/heat, arming the security perimeter, or whatever.
I think the reason is that voice recognition is hard and so far only the "BIGASS TECH!!!" corporations have been able to make it "mom or granny ready" — and they have no incentive to do that for free and let us make our own mash ups. They want to wall us into their ecosystems.
So from that standpoint, this looks pretty cool to me — even if the voice recognition isn't as good as the big three.
OTOH, to rebut my own point: I got the new Apple Watch Ultra and I noticed that I can map the side button to a "shortcut" (the Apple term for a script you create yourself to automate something) that just transcribes whatever I say, and sends it as text over SSH to any host I want. On my local LAN, the delivery time is well under 1000ms.
So that's getting pretty close to being able to use Siri as a generic voice recognizer, and then piping the input into whatever arbitrary/homebrew system I want.
To do it purely with voice though you have to be like "Hey Siri, do the funky chicken" (after naming the shortcut "do the funky chicken"). And then say the actual command phrase you want your home automation to do.
This is not a viable way to treat your original, most ebthusiatic customers. They will go on forums like HN and bitterly complain, warning other potential customers not to invest in a company that clearly does not respect its users.
[0] https://www.kickstarter.com/projects/aiforeveryone/mycroft-m...
¯\_(ツ)_/¯
If you've got an iPhone... put it in to airplane mode so that it is local only. You'll note that Siri no longer works when you do this. However... open up the notes app and tap the microphone. Do some interesting text...
> Mister Smith said that he wanted a two by four and half of a pie.
(if you don't have an iDevice, it transcribes this as:
> Mr. Smith said he wanted a 2 x 4 and 1/2 of a pie
That is without a network and done in real time. We can compare the relative processing capabilities of an iPhone and the RPi, but offline speech to text is feasible on a device of limited capabilities.
Additionally, you can do a limited vocabulary speech to text on chip ( https://www.imagesco.com/articles/hm2007/SpeechRecognitionTu... - https://www.amazon.com/HM2007-Speech-Recognition-Integrated-... ). This can handle the specific incantation common tasks (think closer to how a car voice control works - say exactly these words in this order), but that can help with performance for things that are often done.
A web search finds this 4-mike array for $63, https://www.robotshop.com/en/seeedstudio-respeaker-mic-array...
Google had an audio dev kits for schools based on their TPU and RPi, https://aiyprojects.withgoogle.com/voice/
I just don't see this being worth the money. Hundreds of $ to make switching music slightly more convenient just seems like a colossal waste of money to me.
For example, I noticed a couple years ago at the store that a regular featureless analog wall clock was more expensive than an Echo Dot
These guys need to pay for their software dev out of hardware sales and can't hope for a runaway success yet, so of course it'll cost a painfully lot more
Maybe once a month it freezes and it just needs to be restarted, anyone can do that who can get used to talking to a robot in the first place.
The more complicated stuff, I agree is not fit for people who aren't comfortable with a terminal. I also use it to control my lights and play spotify, but I'm only able to do that because I'm comfortable messing around with it and have the skills (and desire) to debug it when it breaks every other week.
It's nowhere near as polished as Alexa or similar, but it's good for the basics or for hobbyists who don't want to be spied on.
What does this mean? Mycroft prefers taking orders from dudes?
The sexism part is horse manure though. A strong accusation on a likely small team on tight deadlines and budgets who cannot cater to everyone all at once, like Big Tech and their massive resources and teams.
OR
$90 for echo show 8" which does many more things. *Including government surveillance.
I wonder how many people have preordered.
My google home's are terrible at this: often one in another room responds, but at least it's only one. When I tried to run Genie (https://genie.stanford.edu/) I had multiple devices responding simultaneously. It was a disaster.
For me, this is the core feature that will enable me to swap out my corporate listening devices for an opensource, cloud-free alternative.
I have come to despise the Google/Alexa/Siri devices. I'll explain why.
I hate that Google Home devices always give you a direct answer when you ask a question. The reason I hate this is because my kids use it to get an answer, without any work, without any thinking, without any consideration that there might be context to the answer. If they were to research and read about the question, they would learn so much more. But, they, like all people, want a simple and compact answer. And, I'm sure Google engineers have their RSUs tied to some KPI that says "make answers as simple and compact" so it will never come out any way other than this from Google.
I hate that Google permits my kids to play the same damn song over and over again. (Cue sentimental music...). In my day, we listened to the radio and it might have been bad for my dad for five minutes and he scowled the whole way as enjoyed some utterly awful pop song, but then that song ended and he didn't have to listen to it for a few hours. Modern radio is worse, but at least you can take a break for an hour before they play (and are paid to play) the same song over and over.
I hate the surveillance aspect of Google. I don't want to have profiles generated of my kids such that when Google revenues dip in a few years they are enticed by an offer from that shady insurance conglomerate that really wants to know whether any of them discussed depression or racism.
So, if I can use a Mycroft device to:
* Permit them to ask questions, but give them answers in a way they have to dig and think and explore, that would be really cool. I'm sure this isn't easy, but it will never happen with Google/Alexa/Siri because they only care about MONETIZING those interactions.
* Give me more control over how media is consumed. The people working at YouTube will never have a KPI for "make sure you can only play one song per hour" and Google Home will never have that KPI, so it will never happen. That will never be something they can MONETIZE. It seems like it will be a lot more challenging to get music onto my Mycroft, but I prefer to play Jazz radio and because there still are live streams, I think you could get off the YouTube/Spotify/Amazon music train anyway. I got rid of so much of my music, but you can play shared files: https://mycroft-ai.gitbook.io/mark-ii/basic-commands#jukebox
* Forget the worries about surveillance. Mycroft right now uses Google for text to speech, but it can anonymize it enough for me not to worry as much.It's VC funded, the VCs are going to want to get a RoI.
And we all know where that leads...
Most people only use the voice assistants for a few simple tasks, which is perfect for an open source project like mycroft. It is, however, very, very bad for Amazon and Google, because those tasks don't make them money. That's why they're all going so aggressive on "you asked for the time, but by the way here's a 5 minute speech on all the easily monetizable tasks I can do instead"
People like the idea of voice assistants, but by and large they don't like all the problems associated with a voice assistant run by Amazon, Google, and Microsoft.
I've actually pre-ordered the Mycroft Mark 2, although no chance to evaluate it yet.
I'm very interested in devices that can do this locally.
I'm not interested in Alexa/Google home anymore AT ALL - I've gone that route, they both work, but they want my dollars all the time, and it's become increasingly clear that if they can't get me making purchases through those devices - they will kill them off, or become ever more scummy in the attempt (Alexa is now including ads in the "did you know" section - "did you know" was already a fucking terrible decision to include, since it's going to marginally increase interaction at the expense of huge user dissatisfaction. But putting ads there has made me leave.)
So basically - I think if anything, we're seeing a speed run of 90s/2000s tech company boom/bust. A huge amount of money poured into the space with no real idea of how to sustainably profit, but the space itself doesn't feel like it's going anywhere.
It's really, really compelling to allow voice control in all sorts of interactions - but it needs to be very clearly working with me, and not trying to subvert my intent for profit. That might even mean it needs to fall back to something like "if this command, then that action" style usage. No more changing commands, no more bullshit ads, no more subversion of what I'm asking it to do.
It needs to obey me, not google or amazon. Otherwise it's a sales rep and not a digital assistant.
I'll certainly grant that... but the price point where Mycroft is, is certainly not near what I'd pay for doing those few simple tasks.
Apple is at the upper end of what I'd spend for such a device (the HomePod mini is $99) - and that's because I'm fairly invested into the Apple ecosystem and thus it can make use of the iTunes library, home automation, calendar items, etc...
If I wasn't invested in Apple, then none of the home assistants other than Amazon (because of the price point for the echo) would be particularly interesting.
I've got a echo show - because its a very nice simple clock/weather interface (that's got Alexa behind it) too (I really liked the Ambient 7 day weather clock when it was available). I've got an echo wall clock that is paired with the echo in the kitchen - it makes timers nicely visible (a sibling of mine has an echo wall clock because its an analog dial that doesn't have any sound with it).
The problems with Alexa of suggesting by the way ("Alexa, stop by the way" - give it a try and yes, it is routineable) are tolerable for how much I'm paying for them and the functionality that I use it for.
This is a word-for-word description of how Siri originally functioned. "You asked for the top 5 romantic resturaunts nearby; here are the top results from Google Search:"
Apple isn't necessarily in the same boat. Siri isn't particularly good, but it does all those things well. Most importantly, it keeps people on iPhones and in the Apple ecosystem, which does make money.
You already have the phone and probably have a device that works with HomeKit so why not try it out. Next, you buy some new lights. Before you know it, you're controlling most of the lights in your house, streaming Apple Music and setting kitchen timers from your Apple Watch. Next time you need a new phone, you're not even going to think about anything else because if you change you won't be able to turn on your lights anymore.
Apple has a plan that works and Amazon doesn't.
Also importantly, Siri mostly runs off your iPhones processor, and apple doesn’t have to pay a big cloud bill for it, unlike Alexa.
I'd contend that it's absolutely not a solution in search of a problem; it's much more of an unsolved problem, and a big part of the "why" is
- voice recognition/assistance tech still maturing
- major players are insisting that the tech supports their walled gardens
- price points are still a problem
The last two creates a conundrum: a lot of times tech prices come down by selling expensive stuff to rich people until the hardware becomes commoditized. But for a good voice assistant, you need a lot of up-front investment at scale. Unfortunately, the companies that are able to do this are also controlling the hardware that can use it, which limits its ability to spread and be useful.
This is why I think Mycroft is important to support:
1. If you can make voice assistant software open-source and plug-and-play, then it frees people up to tinker with form factors
2. Part of Mycroft's pitch to businesses is that they can make custom solutions. There are probably a thousand big businesses that might want to get into this space but don't want to rely on Amazon because they want to control the experience and not give up their data. Maybe Target wants to stick virtual assistants around their stores, or maybe a hospital wants to give tools for surgeons.
I also think there's an opportunity for voice control in home stereo, where someone decouples the speakers from everything else. It's still annoying to work with Bluetooth in 2022, and Sonos is still pricey, and another walled garden. I'd love to have a simple controller that connects a dumb speaker to Wi-Fi and lets me voice-control it to play music from a library of my choosing. That's not a thing yet, right?
Especially interesting as voice recognition is much easier, cheaper and more efficient within the limited space of a specific usage.
For what it's worth, Google Assistant does have an open API to create new devices. It's not open source, but you can certainly experiment with your own custom form factors. There's even a tutorial:
https://medium.com/google-cloud/how-to-build-your-own-smart-...
Really? An Echo Dot is $25 on Amazon right now. Which, if you use it at all, is pretty reasonable. (To be sure, if I were using it for music to any degree, I'd probably get a model with better speakers.)
For music, I have an old phone connected to a stereo receiver. So it has voice control although I mostly pick a playlist or album manually.
Those companies are failing to profit because they don't understand that a digital assistant needs to be working with me, locally, and not subverting my intent.
It just needs to be my device, and not a sales rep for google/amazon. I use voice controls all the time at home - it's astoundingly useful in all sorts of situations, and I'm not even disabled (where it's literally life changing in some cases).
Unfortunately, Mycroft is not very open itself. Sure, most of the code is open and available, but I tried to contribute and found my PRs ignored for weeks. When they were finally ready to merge them, their poor response cause me to lose interest in the project. At that time, they did not seem interested in cultivating a strong developer community around their core technology components; they were doing their thing, and they wanted the community to implement “skills”. I got the impression that community could either get on board or stand aside and watch them work. For that reason alone, I feel fairly certain that this project will fail eventually as well, and their hardware will become yet another high-tech relic of a paperweight.
As a formerly enthusiastic kickstarter backer, I cannot recommend the Mycroft project as the basis for a product; you don’t own and can’t control the platform on any meaningful way (short of forking it). It might be a better choice than a closed platform, but not enough to make me want to put any money in it.
- controlling smart devices (thermostats, TVs, speakers)
- broadcasting messages
- reminders / tasks
- asking questions
However, none of these use cases generate any revenue for Google afaik.
That set of functionality alone makes them well worth the money for us.
Maybe you meant it like, "the voice assistant space isn't going to generate huge profits, and thus giant corporations will lose interest".
But even that is absurd. They will still have to do it as a loss leader. Maybe not Amazon — because they just ship us our toilet paper and protein bars and shit. They don't have an "ecosystem" (although they gave it a halfhearted try a few times).
But the chance that in 2032 people just like... don't have voice assistants? It's literally zero, barring an actual WWIII cataclysm reversion-to-barbarism event.
> doesn't work as well as people were hoping
Nothing does, until it does...
> little monetization
Yep, that might be right. But it doesn't necessarily mean the space is "dying". Just that it might not be amenable to oligopolization.
Theres a lot of things its really not good for and people have tried them all I'm sure. But where it fails the hardest is being able to increase sales volume for Amazon or increase ad revenue for Google - the only path to monetization seems to be to force it in - and THAT is what is dying.
And who the heck wants it in a ring or glasses??
On the other hand, Mycrodt sounds like something people would actually want to use provided that it can operate locally and doesn't send any data outside the home.
I think the story that you read simply says that it’s hard to monetize. You are inferring more than what the story says.
Voice assistants are here to stay.
I eagerly await the day when I can simply say respond to this post then begin writing with my voice.
I used to use a text-based assistant service called I want Sandy and it was great. Then Twitter bought the company and they went away.
Siri has this as an accessibility feature since iOS 11, but might not be exactly what you're looking for
I was pretty excited about it. I bought a ReSpeaker 2.0, which is an embedded device that can run Linux and has a six microphone array. I designed a custom 3d-printed case to hold the ReSpeaker and a small speaker to make my own little "Jarvis" box (Iron-man reference).
My favorite part about the whole thing was the customization. I wrote a couple of skills to do some other things for me. For example, I could say "Where can I watch X?" and it would use an API to search for a TV show or movie to see where it was available on Netflix, Amazon Prime, Disney+, etc and let me know. It's always been annoying to go Google and try to figure out where I can watch something streaming online, but limited to only the services I currently subscribe to. I wrote another skill that tied into my couchpotato instance so I could say "Download the movie X" and it would go find it and download it. If it found multiple matches, it would read off the top few matches and let me choose the correct one. I even tied those skills together so if the first skill couldn't find a movie at one of my streaming services it would ask if I wanted to download it and I could simply say "yes". I also modified the code to use a custom text to speech API so I could configure Mycroft to use a custom voice.
It was all really cool and I had a lot of fun playing with it. The biggest problem I ran into was the wake word recognition. It worked mostly OK for me on the ReSpeaker from close range but I found as I moved away it went downhill. It was especially bad if I had my device playing music, which is possibly the most common thing I was using my Google Home mini for. I had hoped that the ReSpeaker would help with this, because it had the six microphone array and some built-in loopback hardware to try and cancel out any noise that that was being generated by the ReSpeaker. So any sound output to the speakers would be looped back into the ReSpeaker and could be subtracted from the microphone's input. I found that I just couldn't get it to work well, though. I think the music was causing vibrations that were overloading the microphone array and causing it to be unable to hear me through the music. It's possible it could be improved with a better hardware design to help reduce vibration caused by the device's own speaker. Maybe it works better now, two years later. I think I had configured Mycroft to use Snowboy for wake-word recognition so I could name my Mycroft something else (Jarvis).
One day the Mycroft installation just stopped working on my device after I hadn't touched it in a week or more and I never went back to figure out what was wrong. It's still sitting on the corner of my desk unplugged. If I could have got the wake-word recognition working reliably with music playing I think I would have used it a lot, but I wasn't able to at the time.
I just recently bought a smart watch with a built in "Alexa" app that allows you to send voice commands to your phone which get processed through the watch's official app. I'm instead using Gadgetbridge on Android to interface to the watch. Some kind hacker updated Gadgetbridge to add very basic support for my watch's microphone, allowing you to send the raw voice data to an external application. I'm hoping I'll be able to use this to revive my Mycroft instance and I'll just send voice commands to Mycroft from my watch/phone via a custom Android app/service. In theory, I'll be wearing the watch all the time anyway and having the microphone on my person and right next to my face should hopefully help with the speech-to-text and I won't have to worry about a wake word at all. I've only just barely started working on this, though.
check their free stuff: https://picovoice.ai/pricing/
We do, it's called typing things into Google.
The wakeword ("hey Mycroft") is done on-device, but everything you say after that is sent to a speech-to-text API. That text is then routed to the appropriate skill to handle. So when you're writing the skill you only worry about the content of that text
https://mycroft-ai.gitbook.io/docs/mycroft-technologies/over...
It's an interesting market that I don't think either has figured out too well. Anecdotally, I've not really seen them used for shopping or even really shopping lists and the search model doesn't seem quite as lucrative as if someone used their phone or computer to search.
The one thing they do seem to do well in is as an introduction to - and hub for - the smarthome but I struggle to see how that will make these big subsides viable.
Which brings us back to this. A bit ahead of it's time perhaps but if Amazon/Google pull back their subsidies then this kind of thing might be where we have to go. I'd be happier not using Google Home and have mostly moved to zigbee switches with home assistant to control now anyway. Maybe voice control was a bit of a flash in the pan?
They need to have some sort of version existing to compete with Siri, and can also shoehorn integrations that wouldn't quite work for Amazon
Were not for this[0] ever expanding page, Assistant seems like the more reliable option
However, not wanting to spend time and money integrating the hardware and the software and building an enclosure around it I'd say it's still a fair deal.
Amazon is merely selling devices at cost, which is why they're losing $10 billion a year on Alexa. All this is probably going to stop soon so people end up with useless devices. Hopefully hackers will be able to run MyCroft on them.
>4.3″ IPS wide-viewing angle, full color, touchsceen
Were you able to find more information somewhere else about it being crt?
Personally I do like the design but I'm quite fond of Fallout/Alien style retro displays so YMMV. As a personal assistant type thing the original vertical prototype with eyes seemed better though.
So does StackOverflow. Teaching people how to do things instead of giving them code ready to copy paste is frowned upon there as well.
Eh, idk if I want to spend $350 on what is arguably a half baked, subpar Google-Assistnat/Alexa.
On top of that, the Echo products are just not well built, not well thought out, and have NOT been upgraded to make them better: They update, but with NO visible benefit to the owner. One example: The Echo Show 8 Cannot and will not keep its display off all night, even if you explicitly command "display off" before going to bed (yes, it does understand and temporarily obey this command!) But sometime during the night, something will wake it up, and the damn thing turns into a lighthouse in your bedroom, waking one of us up.
I'd really like to find Alexa more useful, but like most folks I know with one, it's mostly just useful as a glorified voice-controlled radio - I'd use it more to control lights and such, if I could get the damn thing to actually realize waht lights are in what rooms, and that dimmer switches and smart lights can indeed share a location that should be controlled together. (Yes, this is supposed to work, but it doesn't...)
I would pay $500 to outfit the house with a central voice recognition processor that would be capable of supporting a dozen or so very secure listeners on the local LAN. Mycroft isn't that solution.
So cold comfort since it’s annoying as hell, but it slowly learns you don’t like it and will back off its frequency. Amazon unsurprisingly tracks “dissatisfaction” responses and adapts rate of things (globally and individually) so you do actually have to cuss out Alexa to change it. It’s slow because obviously it’s profitable but it does happen.
I'm very grateful to the Mycroft team because I love smart speakers but am not willing to sacrifice my privacy to such a degree as I would have to to use a google home or Alexa or anything like that. That does not mean I won't point out its flaws.
I'm saying, new business idea: give up on the old strategy of Alexa driving induced profit in other business units, instead charge more than the breakeven cost of building devices and running the service. This will likely be substantially more, and fewer Alexas will be sold. But the users that do actually get a lot of value from the device and service they are using will pay more for it.
The product/market fit to test is: How many customers would pay more for a device that's not trying to sell you things? Can you get a solid (but smaller) business just by charging more for the device? What features would you add to persuade the marginal user to pay more? Offline mode for privacy-conscious users? Lean in to home automation features? What about a true AI-powered personal assistant? What about per-user language training with a local model that gets refined by your voice samples, with that data not shared back to the cloud? Etc.
Simple startup-style product iteration stuff here. You had a customer hypothesis and a growth model, and it was proven unviable. So can you pivot to find a viable business?
If you need to go to a specific spot in the house to write some text that turns on a light it seems easier to just walk to an actual light switch? For general automation then I think there are some visual block-based configurators to set up triggers for smart appliances otherwise.
This is not true anymore. Latest iPhone models have offline Siri working to some extent
What's more... consider https://mycroft-ai.gitbook.io/docs/using-mycroft-ai/customiz...
> In order to provide an additional layer of privacy for our users, we proxy all STT requests through Mycroft's servers. This prevents Google's service from profiling Mycroft users or connecting voice recordings to their identities.
If Mycoft goes offline or doesn't pay their STT bill ( https://cloud.google.com/speech-to-text ), the device loses a bit of its functionality.
When my sister visits, it almost never responded to her for example.
And in my personal experience, I used to have a very deep voice and it always responded to me. I decided I would rather have a more androgynous voice and did some voice training to accomplish that and use a pretty neutral pitch. Now it only responds to my normal voice about 50% of the time, so I intentionally drop my voice an octave whenever I speak to it.
But somehow it's not just about pitch either. I have friends who are trans men, and speak in a deep voice but they still have trouble getting Mycroft to respond.
They have instructions on how to train your own version of the wakeword listener.
https://github.com/MycroftAI/mycroft-precise#train-your-own-...
I wonder if the Mycroft project people are aware of the issue?
Although it hasn't gotten better from my perspective. But until I searched for this, I was totally unaware they've provided an easy way to train your own model for the wakeword! https://mycroft-ai.gitbook.io/docs/using-mycroft-ai/customiz...
I will have to try that and if I remember I'll update here if it improves my situation in a few days
i was going to add picovoice's speech to text with rasa https://rasa.com/ but i didnt have time, will give it a try over the holidays.
I see your point but not every open source project lives forever, like sonos killed snips.
Yup, price point is solved for the "just timers and some music" people, but now you're stuck in Amazon's orbit.
https://epic.org/wp-content/uploads/privacy/internet/ftc/EPI...
EDIT: other point I forgot - original article says they're bleeding money, so clearly $25 isn't sustainable right now anyways.
One of the bits in there caught my eye...
> We created a language-specific phonetic specification of the "Hey Siri" phrase. In US English, we had two variants, with different first vowels in "Siri"—one as in "serious" and the other as in "Syria." We also tried to cope with a short break between the two words, especially as the phrase is often written with a comma: "Hey, Siri." Each phonetic symbol results in three speech sound classes (beginning, middle and end) each of which has its own output from the acoustic model.
And the British version getting false positives on wake up with world politics.
The specifics of the wake up and that its done with a ML model rather than a low power wake word chip akin to https://www.syntiant.com/post/syntiant-low-power-wake-word-s... is also interesting - and impressive that they were able to get it to be that low power.
I got an echo (alexa) for free and use it for home assistant. It only works when I have an internet connection. So when my internet is out, I cannot turn my lights on/off with it. I understand why, but i too would REALLY like to just have all functionality dependencies for home automation to be local.
While an improvement I don't want a cloud-dependent system except where it's unavoidable: like having it read me news.
Can a raspberry pi handle the server functionality I wonder?
I personally don't think there is enough cybernetica to control with voice. At some point there may be, but right now, the internet is just one giant consumption stream with a few searches and purchases now and then.
That digital daemon experience taught me that I care more about physical intelligence than verbal intelligence when it comes to my technology. I'm verbally intelligent myself, I don't need an AI who can't even speak correctly, let alone understand me, be my verbal interface to the world.
There is no way that a device can meaningfully parse information from the internet and present it to you right now. At some point? Sure maybe. But it's definitely not now.
What I want, and what I was pretty clear about in my list of use cases is the ability to push a button (or run a function with parameters) with my voice.
My exact use cases
- Kitchen timer: StartTimerFor(duration, timerName?), HowMuchTimeOnTimer(timerName)
- Alarms: SetAlarmFor(time, alarmName?), WhatAlarmsAreSet(), CancelAlarm(alarmName | time)
- Lights/devices(ex: tvs, ACs, etc): TurnOff(DeviceName | RoomName | ALL), TurnOn(DeviceName | RoomName | ALL)
- Conversions: WhatIs(numberOfUnits, inSecondUnit)
- Current time: WhatTimeIsIt(inlocation?)
- Current weather (optional, I don't use this a ton and it requires an upstream source, although HomeAssistant already gets that info for me): WhatWeatherIsIt(atTime?)
---
I've found basically all of the voice assistants I interact with are actually really damn good about understanding roughly those patterns (Alexa was the best, but only for the first year or so [honestly - it was actually wonderful as a beta product] - it's gone markedly downhill over the last several years as they try to cram in more detection and more features).
They just insist on trying to sell me on other parts of the experience (check out this tv show, there's a sale on, use this product, did you know this? did you know that? etc). And I don't want it.
But I'm more than happy to pay a fair sum of money for a thing that will just reliably do those commands.
It's incredibly liberating to be able to do those things with my hands busy, or my eyes closed, or while lying down.
Tack on an HDMI port or a display so I can view a recipe and I will literally give you money right now for this thing (and I have - since I've tried most commercial voice assistants).
Long term - I'll probably end up just cobbling together my own version using Rhasspy/Mycroft or another text parser/STT engine, and HomeAssistant or OpenHab if I can't get what I want commercially. I just know that I'll end up spending more money and time on it that way (which is not the end of the world, I just have other hobbies at the moment, and a very young child).
Amazon is selling the devices at or below cost and hoping that Alexa will make the money (which it doesn’t)
Look at all those Korean novelas and shovelware on Netflix. There are cohorts of subscribers who remain highly loyal because they're into it. Netflix isn't necessarily swinging for the fences with high brow, popular content that competes with the best studios in the world. Instead, they pump out a wide variety of content that keeps the maximum number of people subscribed.
Apple is similar. They promote features - whether it's Siri, health, privacy, family sharing/controls, etc. - that will strongly appeal to some cohort and keep them on the platform. Then, they incrementally hook you into services until you're buying $1000 devices for the whole family and paying $30/mo for the services bundle. And once you're there, they have you because the switching cost involves turning your digital life upside down.
I'm sure enough people find some small use for the voice commands that it's a good feature to have on the phones.
Apple isn't trying to make money with Siri. It's using Siri to make its ecosystem of Apple Music and similar more valuable to its customers.
The limits that Apple puts on what it can do makes that cloud side computation less expensive.
---
Consider that bit - less expensive. Apple doesn't run its own cloud in the way that Google, Amazon, or Microsoft do. So what does Alexa cost? It costs for AWS cloud time. That's the expense that it's running. Those skills that people use run on AWS compute time rather than a phone's local cpu and battery.
A google search "costs" about 1 KJ of energy. Alexa has similar costs somewhere just for energy and other costs for the maintenance of the additional software and content. It costs something to maintain that joke database.
I can't even say "Turn on the lights and the fan". I have to present it all in bite-sized chunks. Come on how hard is that.
I have a feeling there is no AI in these supposedly intelligent assistants. Only scripted stuff.
It's not "Siri is smarter" but rather "Apple is working on minimizing cloud costs because it is entirely a cost for them."
Siri is scripted and limited to make it locally "smarter".
With Amazon and Alexa, everything goes to AWS because that's where the entirety of Alexa's processing happens. The hardware devices in the homes are "dumb" terminals for AWS with a voice interface. This allows Amazon to make use of AWS as much as it can and do something with "surplus" computing power that it has available on AWS.
“What’s it like outside right now?” => GetWeatherIntent arguments=[time=current location=device_location]
There’s no Omni intelligent being making API calls anywhere. It’s NLU with a decision tree(s) based on the intent and arguments
When I give a double command Siri literally tells me she can't do two things at the same time and I have to present them as two separate commands.
Perhaps it's because I have it set to UK English? Perhaps it's smarter in US English. I'll have to try that.
Most companies treat you know English as European, and European customers as second in line for new features after America.
Or more specifically, treat American English first for voice control stuff.
https://www.amazon.science/blog/on-device-speech-processing-...
This is also part of the work that Apple has done (and is likely part of the making Siri cost less for cloud compute).
https://www.macrumors.com/how-to/use-siri-offline-ios/
> In iOS 15, Apple moved all Siri speech processing and personalization onto your device, making the virtual assistant more secure and faster at processing requests. This also means Siri can now handle a range of requests entirely offline.
> Once you're using iOS 15, you don't need to enable anything for Siri to work offline. The types of requests that it can handle without phoning home to Apple's servers include the following:
Create and disable timers and alarms.
Launch apps.
Control Apple Music and Podcasts audio playback.
Control system settings including accessibility features, volume, Low Power mode, Airplane mode, and so on.i just wanted to share my 2 cents, as it's not unique to Apple and the cloud can be costly even if you own it. Big tech has been investing in on-device for a while. besides voice commands, apple and google do transcription locally too. because now you can have local speech to text with cloud level accuracy and of all the reasons you shared - cost, privacy, latency etc. (but again, i'm biased)
https://www.androidauthority.com/voice-typing-opinion-322134... https://techcrunch.com/2022/05/17/apple-adds-live-captions-t...