Mycroft II Voice Assistant

[Scottmichaelj]

https://www.kickstarter.com/projects/aiforeveryone/mycroft-mark-ii-the-open-voice-assistant

 

https://mycroft.ai/

 

They talk about “privacy” but they have to be doing something to get the voice recognition working. I am going to look into this project and follow it a bit. The “super backer award” for $25 maybe a good way to speak direct with the team to help get things into the system that can work directly with the ISY.

 

As always just wanted to share. Discussions welcome, let’s stay on topic

[Teken]

https://www.kickstarter.com/projects/aiforeveryone/mycroft-mark-ii-the-open-voice-assistant

 

https://mycroft.ai/

 

They talk about “privacy” but they have to be doing something to get the voice recognition working. I am going to look into this project and follow it a bit. The “super backer award” for $25 maybe a good way to speak direct with the team to help get things into the system that can work directly with the ISY.

 

As always just wanted to share. Discussions welcome, let’s stay on topic

 

What exactly interest you about this piece of hardware?

[Scottmichaelj]

What exactly interest you about this piece of hardware?

Not the hardware more than the privacy, open source and being able to run on third party devices. However the requests have to go somewhere so I want to dig deeper into the TOS and privacy.

[Teken]

Not the hardware more than the privacy, open source and being able to run on third party devices. However the requests have to go somewhere so I want to dig deeper into the TOS and privacy.

 

When I saw the campaign a few days ago the only items that interested me was having the ability to select different wake phrases, LCD. Based on their existing video they have a long ways to go with the sound of their voices. I'm not sure what they mean by AI and may have over looked what they were trying to relay to me. But I didn't see anything that could be considered as AI.

 

AI to me is something that can take new input and learn from that new (unknown) data.

 

As far as I can tell this product calls home to some cloud server and performs a query to reply back with X response. If so this mimics what Alexa / Home does but poorly in my view. Having said this what you called out for having the ability to install this on a RPi is a interesting feature. But Amazon has been doing this for a few years now so its not new or ground breaking.

 

For me what pulls at my pant legs is the whole over use of the phrase AI . . .

 

As of this writing there is no AI in Alexa, Home, and certainly not in this product . . .

 

With the likes of Apple, Amazon, Google, bar none the three most powerful and profitable companies in the world. If none of them has been able to offer true AI there is zero chance a start up company which says their software which runs on a $35.00 RPI is going to give you anything remotely close to AI.

 

Now, if this gizmo can give me something that is capable of simply interactive Julie U.S. that is a whole different ball game for me. If the scope is within that realm and capability I have buy in.

 

Can it?? I don't know . . .

[giesen]

https://www.kickstarter.com/projects/aiforeveryone/mycroft-mark-ii-the-open-voice-assistant

 

https://mycroft.ai/

 

They talk about “privacy” but they have to be doing something to get the voice recognition working. I am going to look into this project and follow it a bit. The “super backer award” for $25 maybe a good way to speak direct with the team to help get things into the system that can work directly with the ISY.

 

As always just wanted to share. Discussions welcome, let’s stay on topic

I've been watching this for a while. I passed on the original Kickstarter but this time 'round the hardware should be much more mature, and the software is maturing quickly as well. I snagged 2 of the early birds

 

Sent from my SM-N9500 using Tapatalk

[giesen]

What exactly interest you about this piece of hardware?

The big difference with Mycroft is that the software is open source, and you can actually run your own "cloud" if you so choose to handle all the voice queries. It's a rapidly maturing project and they have plans to implement a new set of voices and new voice recognition software within a month or two, long before the new hardware launches.

 

You can also write and contribute your own skills (not an easy proposition for Amazon Echo/Google Home).

 

 

Sent from my SM-N9500 using Tapatalk

[Teken]

In the big picture more competition and ideas helps push the voice interactive technology and other companies. I wish the company all the very best in this endeavor but I'll sit this one out.

 

Thus far Alexa has provided the bulk of what I need. If and when she can alert me of (X) conditions without being prompted voice technology at that point has past the milestone all of us have been waiting for.

 

2018 and beyond is truly going to be about shakers and movers.

 

 

Sent from my iPhone using Tapatalk

[asbril]

Does Open Source mean that it will be compatible with both Alexa and Google Home, and therefore with ISY ?  

 

I am not that concerned about Amazon and Google using my data. I already use extensively their services (Amazon Prime, Gmail, Google Voice etc) that one more product won't change much.

 

I would also bet that the December 2018 delivery date will give time to Amazon and Google to come with new generations of their products.

 

What I am looking for now is (an easy way) for ISY to push messages to Google Home, including initiating a phone call. I am also waiting to see what Google will do compete with Alexa's Show and Spot.

 

The Mycroft device has a screen but does not appear to have video streaming or video calling.

 

But I welcome more and new AI products.

[lilyoyo1]

Does Open Source mean that it will be compatible with both Alexa and Google Home, and therefore with ISY ?

 

I am not that concerned about Amazon and Google using my data. I already use extensively their services (Amazon Prime, Gmail, Google Voice etc) that one more product won't change much.

 

I would also bet that the December 2018 delivery date will give time to Amazon and Google to come with new generations of their products.

 

What I am looking for now is (an easy way) for ISY to push messages to Google Home, including initiating a phone call. I am also waiting to see what Google will do compete with Alexa's Show and Spot.

 

The Mycroft device has a screen but does not appear to have video streaming or video calling.

 

But I welcome more and new AI products.

I'm with you on using Google services. It's unfortunate what Google does with my data but at the same time, the benefits (so far) out weighs it's cons. I'm hoping for comes out with a spot competitor soon.

 

The product seems cool from a developer standpoint. However I just don't see any reason to take a chance on them. They haven't given a compelling reason for me to jump on it over Google/amazon.

[Teken]

I'm with you on using Google services. It's unfortunate what Google does with my data but at the same time, the benefits (so far) out weighs it's cons. I'm hoping for comes out with a spot competitor soon.

 

The product seems cool from a developer standpoint. However I just don't see any reason to take a chance on them. They haven't given a compelling reason for me to jump on it over Google/amazon.

 

I'll give the new kids a break as noted more competition in this realm is more than welcomed. Having said this, they will need to push this product to do a lot better. The form factor isn't something I am keen on never mind the mickey mouse voices in use now. As noted up above what interests me is (IF) this device can actually initiate speech with out me ever prompting it to do so.

 

I'll wager Amazon will have that portion hammered out by 2018 long before this product ever will.

[asbril]

I'll give the new kids a break as noted more competition in this realm is more than welcomed. Having said this, they will need to push this product to do a lot better. The form factor isn't something I am keen on never mind the mickey mouse voices in use now. As noted up above what interests me is (IF) this device can actually initiate speech with out me ever prompting it to do so.

 

I'll wager Amazon will have that portion hammered out by 2018 long before this product ever will.

Teken that is one of my points as well. We are in January and a December delivery (if they make it) is too far away. Like you, I bet that Alexa and Google Home will have updates before then.

[Teken]

Teken that is one of my points as well. We are in January and a December delivery (if they make it) is too far away. Like you, I bet that Alexa and Google Home will have updates before then.

 

I'll wager they won't make the December 2018 dead line most projects rarely do when its a complex device. In all of the crowd funding projects I backed out of the 45 something less than half (47%) delivered on time per their Advert. For me it didn't really matter if they made the self imposed dead line so much as the product was solid and the features and capabilities they advertised to me came true.

 

The only benefit of getting on the ground floor is receiving the product first and a possible break on the price vs retail.

 

Having a direct hand to help the company develop an idea into a physical piece of hardware is truly a great experience.  

[Scottmichaelj]

So I am digging into this company and found on their blog a post about in March how they are going to move to "DeepSpeech" which is already available if you want to install it on your own hardware.

 

https://mycroft.ai/blog/mycroft-speech-to-text-and-balance/

 

I cant find "real" non-marketing hype about privacy outside:

 

"Here at Mycroft we take privacy seriously. We don't preserve data unless we are given explicit permission. We don't sell data to third parties and we don't intrude on the lives our our customers. We aren't trying to sell you products, dominate online advertising or own your digital identity. Switching to DeepSpeech advances our goal of providing our users with the highest qualtiy, most private experience possible. We're proud of our commitment to privacy and look forward to bringing it to more and more of the Internet over time."

 

This blog post is pretty interesting though and overall gives you a sense of their "philosophy", but how that transitions into business is a different story. For example, what info do they keep? How long is their data retention? How will they process a warrant from a government agency? Things of this nature is what I want to know. I cant find a answer to these questions. I may drop them an email or call their "media" line to see if they can direct me to someone who can.

 

I did like that "Users can easily select Google, Watson, or Wit.ai as a STT provider (or Kaldi, Bing, Houndify or more if you are willing to get your fingers dirty)." IBMs Watson might be a great STT since I know IBM is more privacy conscious.

 

I will post more as I find.

 

Side note they are also working on Machine Learning & Open Source Speech to Text with Mozilla whom is also doing some great things to protect privacy.

 

https://research.mozilla.org/machine-learning/

 

Here is another Mycroft Blog post about privacy.

 

https://mycroft.ai/blog/privacy-machine-learning-open-data-set-opt-feature/

 

Since the software appears to be open source you can actually opt out of them collecting data. What I am not so clear is if this means the software is strong enough on its own hardware to do the actual TTS computing.

[Bumbershoot]

Does Open Source mean that it will be compatible with both Alexa and Google Home, and therefore with ISY ?  

 

 

No, not at all.

 

The term "Open Source" generally refers to the license that the code is published under.  There are many examples of open sourced code, such as Linux (GNU General Public License), BSD Unix (BSD Licenses), Firefox (Mozilla Public License) and even macOS (Apple Public Source License).  This can be of great interest to folks who would like to keep software project going in case the original developer closes down development, among a great many other things.  An example where this happened is with the iOS app called Locative, which many on this board use to create geofence applications.  As I understand it, the developer moved on, but made the code available under an open source license so others could carry on with it.

 

Here's the license Locative is published under, as a very simple example:

The Geofancy License

Copyright (c) 2013-today Marcus Kida

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), 
to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, 
and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

You are not eligible to distribute this Software under the name or appearance of Geofancy, you may release it under another name and appearance though.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

[giesen]

So I am digging into this company and found on their blog a post about in March how they are going to move to "DeepSpeech" which is already available if you want to install it on your own hardware.

 

https://mycroft.ai/blog/mycroft-speech-to-text-and-balance/

 

I cant find "real" non-marketing hype about privacy outside:

 

"Here at Mycroft we take privacy seriously. We don't preserve data unless we are given explicit permission. We don't sell data to third parties and we don't intrude on the lives our our customers. We aren't trying to sell you products, dominate online advertising or own your digital identity. Switching to DeepSpeech advances our goal of providing our users with the highest qualtiy, most private experience possible. We're proud of our commitment to privacy and look forward to bringing it to more and more of the Internet over time."

 

This blog post is pretty interesting though and overall gives you a sense of their "philosophy", but how that transitions into business is a different story. For example, what info do they keep? How long is their data retention? How will they process a warrant from a government agency? Things of this nature is what I want to know. I cant find a answer to these questions. I may drop them an email or call their "media" line to see if they can direct me to someone who can.

 

I did like that "Users can easily select Google, Watson, or Wit.ai as a STT provider (or Kaldi, Bing, Houndify or more if you are willing to get your fingers dirty)." IBMs Watson might be a great STT since I know IBM is more privacy conscious.

 

I will post more as I find.

 

Side note they are also working on Machine Learning & Open Source Speech to Text with Mozilla whom is also doing some great things to protect privacy.

 

https://research.mozilla.org/machine-learning/

 

Here is another Mycroft Blog post about privacy.

 

https://mycroft.ai/blog/privacy-machine-learning-open-data-set-opt-feature/

 

Since the software appears to be open source you can actually opt out of them collecting data. What I am not so clear is if this means the software is strong enough on its own hardware to do the actual TTS computing.

 

 

The hardware only does STT for the wake word. The rest is done on a backend system, either at Mycroft or one that you host yourself. The current iteration aggregates queries from all Mycroft users and sends them to Google Voice (or which ever service is used), in theory eliminating your queries from being personally identifiable. Once DeepSpeech is launched, the voice processing will be done directly at Mycroft (or at your home if you host your own server). You can talk to most of the people at Mycroft at https://chat.mycroft.ai. It's regularly attended by the CxO's, as well as the hardware and software developers. I'm sure they'd be more than happy to answer your questions.

 

Note that all of the Skills run directly on-device (as far as I know), the only thing that is done via the Mycroft cloud is speech processing.

[giesen]

These are the hardware specs of the Mark II for those interested (a huge jump ahead of the RPi-based Mark I):

 

Xilinx quad-core processor

Far-field 6-microphone array

Hardware AEC, beamforming and noise reduction

Stereo sound with dual 2" drivers (10W)

4" IPS LCD touchscreen USB Type A

MicroSD card slot

Bluetooth in

Wifi

3.5mm audio out

18W power supply with international adapters

 

The Mark I was very focused around being open hardware that anyone can build, but it had some limitations (especially around the microphone, which quite frankly, sucked). The Mark II is a little more closed hardware-wise (they may not be able to publish all the hardware diagrams for all the components). They of course strive to be as open as possible, but commercial realities limit some of their options.

 

Note you can also run the software on a Mark I (their legacy device from their first Kickstarter compaign), build your own RPi-based solution, run it on Linux, or run a partial instance on Android (which still has to connect to mycroft-core running somewhere else, like an RPi).

[Scottmichaelj]

The hardware only does STT for the wake word. The rest is done on a backend system, either at Mycroft or one that you host yourself...

Thanks for the information.

[Teken]

Just to quickly summarize no matter what hardware is in use (RPi / Other) this device requires a constant Internet connection to operate, yes?!? If so, this will not replace my current stand alone (local) Julie U.S. implementation.   

[giesen]

Just to quickly summarize no matter what hardware is in use (RPi / Other) this device requires a constant Internet connection to operate, yes?!? If so, this will not replace my current stand alone (local) Julie U.S. implementation.

For the moment yes. Once the backend code is polished up, you will be able to run your own backend server (albeit not on the device itself and it will require significant hardware)

 

Technically you can plug in any STT engine you want, so if you had the skills and determination you could plug Julius into Mycroft. From what I've read though, Julius is not exactly a great STT engine (feel free to disagree)

 

Sent from my SM-N9500 using Tapatalk

[Teken]

For the moment yes. Once the backend code is polished up, you will be able to run your own backend server (albeit not on the device itself and it will require significant hardware)

 

Technically you can plug in any STT engine you want, so if you had the skills and determination you could plug Julius into Mycroft. From what I've read though, Julius is not exactly a great STT engine (feel free to disagree)

 

Sent from my SM-N9500 using Tapatalk

 

Julie U.S. requires complete and total user setup up and planning for sure. But once set up the system just operates with out any further user intervention or maintenance. My hopes was this new piece of hardware was capable of operating in a local stand alone mode but obviously given the low price and current hardware its not realistically possible.

 

The benefits of Julie U.S. is its stand alone with out the need of any tethered Internet connection. Because its isolated and local there is zero chance of privacy / security issues with this type of set up. The short coming to Julie U.S. is obviously the feedback is by its very nature canned and defined. But the reality is once the basics to advanced conditions have been defined there is very little reason for new vocal prompts / feed back.

 

As of this writing the benefit of using Julie U.S. and Blueman's method to allow Alexa to reply back with real time values is I have the best of both worlds. Ideally if and when Alexa can speak with out being prompted this will make it even better but of course requires a constant tethered Internet connection.

 

But, based on how Amazon has carried themselves privacy & security is less of concern when speaking about voice integration ~ well at least for now.