What do you want to see from Mycroft?

Great feedback @neurosisone, and we totally agree with you!

We currently prototyping something like this internally that will help ‘train’ Mycroft on incorrect answers. When it’s release, we’d really love your help in training it!

1 Like

I’d be more than happy to help! :slight_smile:

And it would give people bragging rights. Could you imagine all the Easter eggs like hey Mycroft why am I so cool. And have several different answers rather than just one such as. I think you’re so cool because you’re talking to me and I’m pretty cool myself Mycroft out. hahaha

1 Like

What do I want see?
I want to see a full implementation of Mycroft that can be online or offline (we choose). I asked about a year ago if Mycroft could be offline and the answer was it was being worked. I asked again a couple of months and still it was not setup to do it yet. You may ask why this is so important.

  1. Let me point you to this:
    Siri/Alexa/Google able to hack with hidden audio

  2. If your house is home automated and your electricity goes out you should still be able to turn on and off the lights and important things around the house. My son (high functioning Autism) has an Alexa and this was the first thing he hated about it when our internet went out and he could not turn on his light (we have since fixed this my putting the lights on a local mosquitto server setup)

  3. If we want to stay private we should have that option while still being able to be online when needed (like weather, local news, etc).

Thanks

Thanks!

1 Like

@krisadair unfortunately I wasn’t able to join in on any of the live sessions that Joshua and Steve held. Are you able to give a summary of what community generated suggestions have made it into the Road Map for 18.08? Thanks.

Hey @aussiew, thanks for your comment. We’re still finalizing the Roadmap for 18.08b, but here’s some of the things that you’re likely to see.

Performance. Now that I have my mark I I notice that there is a lot of delays. Hey mycroft do something. Then a long pause before the okay response.

My degree is in human computer interaction, you get 1/10th or a second to respond. You can probably add a second with a umm but most actions should complete by then. For things where you have to calculate you might get longer, but for the most part mycroft isn’t trying anything where users should accept delays.

I think this probably forces everybody to run everything locally, the lag of crossing the internet is too long even before spt calculations are counted.

Compared to a google home or echo, my mk1 seems about a second or two slower, maybe 2/3 the overall speed of those others. Some skills take longer than others, as well. The mk1/picroft hardware is, in my view, a big part of that. Run on more significant hardware a bunch of the delay gets taken away.

Running locally for everything may not be better, the speech to text engines in the cloud can do better (speed and accuracy) than most general STT installations without high-end gpu’s or other server level resources in my experience.

i think these things are worth able even after 4 years

What I want to see from Mycroft is some really basic and ‘unsexy’ stuff. I don’t yet think this project is at the point of considering ‘next phases and steps.’ I know that sounds cranky and contrarian, but I don’t mean it in a negative way at all! I hope my opinion (since you asked) is seen in a negative way. I mean my critiques to be positive.

I’ve been using Linux since 1998. I’ve seen a bunch of projects come and go! And that’s okay. Most usually the projects go away because they were mostly driven by a singular personality. It was his or her pet project. Sooner or later, it was no longer enjoyable to maintain the project. Supervising the project had become less about creativity and more about routine maintenance. Passion is lost.

I can usually see that turn coming when the leadership starts asking questions like “What else is there to be done with this?” Sometimes the reach out to their community asking “What would you like to see done with this project.”

Sometimes these questions are asked after the founding personalities leave the group, or take a (much needed) hiatus.

The questions aren’t bad. Sometimes they just show the team is either overwhelmed, underwhelmed, or looking for direction after having recently floundered.

Maybe that’s not the case with Mycroft! It is a cool project. But I suppose The allure to ‘monetizing the project’ may not be in the picture. With Mycroft’s two founding members having recently left the organization I’m guessing that’s the case here. Most founders don’t leave before they bring a financially viable product to fruition.

So all that to just lay some groundwork for what I’d like to see out of the project. I’m ‘just an end user’ who enjoys seeing the possibilities of the Mycroft project.

I think more attention needs to be paid to the unsexy foundational fundamentals of the project. It is not yet at the point where masses of people can easily use it or even write for it. It is just not yet adoptable or adaptable as far as I can see. It’s close, but you need to take a few months and get a “down to the basics” task-list together and get the items punched out.

This won’t be done with a “we’re always working on that stuff” mentality. It really needs to be an “all hands on deck” mentality. That way the boring stuff like better documentation and skills standardizations happen with no one feeling like they’re stuck doing it.

I’ve seen a bunch of products die off because the leaders wouldn’t admit that their ideas were untenable. A great example was when years ago a few guys decided to make a KDE tablet. Their were a few enthusiasts who thought the idea was neat. But when it became clear that the project was doomed the leadership wouldn’t give in. Even when other tablets were already made that you could load KDE on already were produced, they wouldn’t give up. Even when their own costs started skyrocketing, they wouldn’t give up. Even when they saw that not many people were willing to support the project with money via preorders (because the price for the product was ridiculously high) … you guessed it… they wouldn’t give up. Eventually, they died off with a whimper, and the founders lost much of the credibility they once held.

I’ve posted elsewhere on the forum that the Mark II is in just such a situation. I’m posting here because I truly want this project to succeed!

You have a good technology. The Mimic3 tech might even be a marketable piece of the puzzle. I don’t know. But your project has not done enough to adequately leverage the popularity and use cases for the SoC (or Raspberry Pi) form factor regardless of the ‘Picroft’ product. You really should dominate that space - or other projects eventually will. A good example of this might be the “OpenOffice Project.” They would not adapt and change to the needs of the users. They pursued a “money model” and eventually were bought by a corporate entity. They were also outpaced by LibreOffice, a fork of their very own project!

/End of rant. :grinning:

4 Likes

OMG that is a well written and honest post which actually is so true IMHO.

I would like to step in for the guys from Mycroft A.I. about the “All hence on deck”.

Before the recorded dev syncs died off, they where actually precisely did iust that. Focus on the core functionality and get it to work perfectly.

To be honest, i have no clue where they are at with it now because of the dev sync silence. However With three FTE developers that should be amazing by now. Perhaps I should try the latest Mark2 dev kit image again to see what they have been up to.

1 Like

I think the community in general is waiting for how the MkII will turn out but really you have x3 very good general purpose python programmers with no embedded engineers or at least they are not being allowed to be so.
So there are still big questions if they can utilise the new audio hat and irrespective of software the results might be confined by design.

There is a huge hole in the ‘opensource’ hardware market as generally much of what is available is near e-waste.
The re-speaker range of devices is low cost but the drivers render most useless with a huge array of bugs that have remained unfixed for over a year.
Currently the channel order of their 4 mic square/linear & 6 mic hat is completely random on record which is such a chocolate kettle I just got a full refund from paypal.
This bug has been so since last September but they continue to sell to anyone foolish enough to purchase.
I purchased with paypal as I know what a terrible awful part of the Seeed company respeaker are.

There is a huge range of relatively low cost audio hardware missing from the Pi eco-sphere and also a complete lack of embedded high performance DSP libs.
Which could be stocked on relatively low investment and developed with little overhead.
So whatever Mycroft do they are always constrained by a quality void at the start of the audio pipeline as it has been now for many years.

For many though waiting for the MkII dev kit is pointless though as they are not for sale.
Mark II Prototype Development Kits are not for sale.

I agree with you about the ReSpeaker stuff. Crap!

The Mark2 (devkit and consumer) does not have any respeaker stuff but is build up around the XVF3510 from the XMOS guys in the UK.

And is in a compete other league compared to anything out there. Said it many times before; Such a shame they do not sell it stand alone as it really fill that gap of a proper voice hat.

Its an upgrade on the Respeaker USB 4 mic which used the XVF3500 but no its very much like quite a range of product.

China comes to the rescue once more if you did want one.

Hopefully its better than the XVF3500 that my Anker power conf had, as was sort of ok but no good in the presence of 3rd party noise unless static constant noise.
It lacks some of the advanced Algs Google use and haven’t heard anything from the devs that might hint of such advances, but will have to see.
How its engineered and implemented is critical and that will only be seen when product finally arrives.

Likely it will be much better than Respeaker though as its there driver support as its non existent and they have no care about shipping hardware with non functional drivers due to kernel upgrades.

“How its engineered and implemented is critical and that will only be seen when product finally arrives.”

FWIW the mk2 dev kit can pick me up across the room speaking at normal volume while playing music full blast

1 Like

No, it’s the other way around.

The ReSpeaker USB version is just the Chinese company once again, grabbing and dumping the minimal required hardware aspects onto a device, grabbing the standard firmware to dump it onto the market without any aftercare or software support. Exactly the same with the xvf3510 board you showed.

XMOS as a (EU) company does support their customers with all the software and driver updates. Their firmware get’s better and better and yes, that comes with a price diffeence compared to all those China boards.

Anyway; We all agree, their is a HUGE need for a proper voice HAT into the embedded world.

But then again, the sj201 hat is open hardware, so who is stopping any one to continue their work and creating an updated stand alone version.

1 Like

I don’t think it is the same as the xvf3510 only has 2x mic inputs whilst the earlier xvf35xx had 4 or more…?

I will wait for what other users say. I never doubted its AEC would work as that is pretty simple stuff nowadays but what I said was ‘no good in the presence of 3rd party noise’ and that was the older gen.

XVF3510-UA for USB accessory devices, and XVF3510-INT for built-in solutions

The number changes are different models and presume newer like the XVF3610 is … ?

Hey there - totally agree on the need for us to focus on the “unsexy stuff”.

It is what we’ve been primarily focused on for the Mark II lately, and whilst we’ve made some good progress it has highlighted a lot of the limitations of the current mycroft-core framework. Rather than putting more patches on top, we really need to look at what foundational pieces would improve these as a whole.

To that end, we’re developing a new development roadmap at the moment and want to get everyone’s help in identifying these “unsexy foundational fundamentals” that need to be addressed. I’ve posted a list of the ones that we’ve already seen and heard here:

But I’m sure there is more…

We really want to get this right before adding lots of bells and whistles on top. An amazing new Mimic 3 TTS is awesome but if we aren’t meeting basic expectations then it won’t grow to what it could become.

So would love any and all thoughts in that thread!

1 Like

I use mycroft with my robot and it generally works very well (only problem really is getting it to reliably find the microphone). I think robotics is an untapped market for mycroft because you don’t see many robots out there that interact through voice and speech. Here’s a link to a tweet showing some of what I’ve done with it.

In order to accomplish this, I had to slightly modify the code initialize the ROS (robot operating system) node in the main function (apparently it has to be initialized on the “main thread” and not in a skill). Obviously, this isn’t something needed for everyone so it’s not something to be incorporated into the main branch, but having a ROS specific branch with that code (and other ROS specific features) could really open opportunities for you. In any event, whatever changes you make, please do not break my ability to use it with my robot.

1 Like

That is an awesome looking project!

It’s also a good highlight of how the current mycroft-core makes it more challenging than it needs to be to integrate Mycroft into/with other projects. I’m not familiar with the framework, but will make sure we keep those types of systems in mind for this roadmap. Thanks :slight_smile:

1 Like