When we started Mycroft the Amazon Echo was in private beta. We’d never heard of it. Google Home didn’t exist either. We started Mycroft because we saw a future where a simple speaker and microphone would become an augmented reality overlay for homes, offices and vehicles. Our thesis was validated – big time – by huge investments from Google, Amazon, Apple, Tencent, Baidu, Alibaba, Samsung and many others. The battle to put smart speakers in the home has become one of the most important battles in Silicon Valley and globally.
For the past four years Mycroft has played the role of the open source alternative to Siri, Alexa, Assistant, Viv and the other proprietary stacks. Our focus has been on building a flexible framework for deploying voice technology. We’ve also spent a great deal of time working to improve the open source community’s solutions to wakeword spotting, natural language understanding, speech to text transcription and voice cloning. All in all we’ve done a pretty good job at executing on these things. Our Precise wakeword spotter works extremely well. When properly populated with data our NLU engine, made up of the Adapt and Padatious intent parsers, are very fast and capable. This engine is also serving more users around the world with our growing multilingual text parsing and formatting library, Lingua Franca. Our latest cloned voice sounds fantastic. On the speech to text side of things we’ve left the heavy lifting to Mozilla who’ve made strong progress, but have yet to ship software that is useful for real world applications.
What we haven’t successfully executed on is shipping a quality smart speaker. We don’t even have a hardware engineer on staff. Our original plan was to outsource this to a company that does have hardware expertise. Our original selection for this was Aaware – a two man startup in Los Gatos. This turned out to be an error. Aaware was unable to deliver working prototypes in the allotted time and when we realized they might never deliver working hardware we pulled the plug and moved on.
Our second effort used off-the-shelf technology from Seeed Studios. We attempted to get the Mycroft stack to work reliably on their Respeaker hardware, but ran into problems with processor time. This board utilized software to handle the Digital Signal Processing (DSP) for the multi-microphone array and basically the hardware didn’t have enough processing power and RAM to run the full Mycroft stack and the DSP.
Our third effort made use of a different Respeaker microphone board from Seeed with a hardware DSP solution and a single board computer from Pine Studios. The Pine H64 is a very capable board with enough processing power and RAM to run the Mycroft stack. Unfortunately, the board does not have mature driver support. As a result we’ve been unable to bring the Mycroft stack up on it reliably.
So we are currently developing on the Raspberry Pi – a platform we’ve been supporting since day one. The Pi is robust, well supported, well documented and affordable. Our design and software team are now focused on a Pi-based Mark II, and we are moving fast and making progress. We aren’t ready to announce a change in hardware, we’ve learned our lesson from the past. We will announce a change when we are confident it is the right decision. However we are ready to show improvements in audio quality, barge-in, and other user experience advances on the Pi-based design knowing that these improvements will carry over to all implementations of Mycroft.
As the CEO of a technology startup I have one primary job – don’t let the company run out of money. Some CEOs handle this by focusing on sales. Revenue funds operations and they spend most of their time working with customers. This hasn’t been possible at Mycroft because we haven’t had a product to ship. The Mark I doesn’t provide a high enough quality experience for larger companies ( retailers, hospitality companies, healthcare companies, banks, etc. ) to deploy our technology to their customers. Our company has been forced into an alternative route: funding operations through investment.
As a result I’ve spent the last four years focusing on raising money to fund Mycroft’s operations. I’ve largely left product development in the hands of our technology team. I’ve been hands off with the product and spent my time making sure our brand is well recognized, our market position is secure and our operations are funded.
Unfortunately companies tend to take on the focus of the founder. What that means at Mycroft is that we’ve focused on marketing and fundraising more than we’ve focused on shipping a quality product. Our technology team has been working on core technologies successfully, but hasn’t created a product that customers can place in their kitchen.
In June it became apparent to me that unless I focus on product rather than fundraising we’d end 2019 in the same place we started it – with a bunch of great core technologies, but without a product we could ship to customers. As a result I made some staffing changes and re-focused the company on creating a quality product.
Over the next six weeks our technology team will be laser focused on taking the great technologies we’ve built and combining them into a customer experience that is on par with what the public got in early versions of Echo or Home. We’ve started doing an internal demo each Friday and that demo has shown us where we need to make improvements.
I’ve stopped fundraising for the first time in four years and am focused on helping the team to put together a quality product. On one hand this is a good thing – as I said above, companies tend to reflect the focus of their founders. On the other hand it means we are not fundraising and for the first time we have a finite runway.
Does this mean we’re refunding our investors? No. We’re using all of the funds from backers along with millions of dollars from investors and, in partnership with hundreds of open source developers, we’re working to make the Mark II a reality.
So what can our community do to help? First off, we have a host of bugs out there that need to be squashed. The community can help by booting up Mycroft on a Raspberry Pi and helping us with critical tasks like stabilizing the various music services, making the network connectivity more reliable, improving the WiFi setup process and adding tagged data to Precise for women, children and non-english speakers.
The other place the community can help – and this is targeted specifically at corporate users or potential corporate users – is by looking into the future and making investments in our technology now. We need investments of time, money and cooperation today, not after the stack is mature and ready for prime time.
Corporate strategists would do well to realize that the type of voice technology we are building is eventually going to replace receptionists, call center representatives, customer service agents and in-store service associates. It is also going to open up new opportunities in healthcare, entertainment, hospitality and product HMIs. The way things are going today this technology – like mobile technology before it – will be the sole property of Google, Apple, Amazon and other huge Silicon Valley conglomerates. If older companies want to have an open option in this space, their strategy teams need to wake up and invest in our effort. The alternative is another Silicon Valley duopoly charging the rest of the world a tax on every transaction, that’s likely to be 30% of your future revenue stream.
So when will the Mark II ship? When it is ready. No one will remember if we ship by some arbitrary calendar date. Everyone will remember if the technology doesn’t work. Our team is hard at work. Believe me, they know we’re late. The community can accelerate the process by squashing bugs, contributing data and participating in our development effort. We’d welcome the help and, as always, appreciate everyone’s support and patience as we solve difficult problems and bring the Mark II to life.
CEO of Mycroft A.I., serial entrepreneur and one of the few entrepreneurs in the United States to build a gigabit fiber network from scratch. Joshua brings more than 15 years of entrepreneurial experience to the Mycroft team.