#StartupsEverywhere profile: Joshua Montgomery, Founder, Mycroft AI
This profile is part of #StartupsEverywhere, an ongoing series highlighting startup leaders in ecosystems across the country. This interview has been edited for length, content, and clarity.
Democratizing Voice Assistant Technologies
Mycroft AI offers open-source voice assistant technology that can be integrated into all types of third-party hardware, from desktop computers to automobiles. We recently spoke with Joshua Montgomery, the Founder of Mycroft AI, to learn more about the need for patent reforms to reduce abusive litigation, the intersection between user privacy and open-source AI development, and how Mycroft AI will contribute to the growing voice assistant market.
Can you tell us a bit about your background and what led you to entrepreneurship and launching Mycroft AI?
I lasted two years in corporate America before I quit and set out on my own. Reflecting on it reminds me of my grandfather’s story as a meat packer in Minnesota. He made the deal to spend thirty years there in exchange for a pension and healthcare. At the time that was valuable. However, now that bargain is gone for most Americans, and stepping out on your own is often the best way to find success. And the government has an important role to play there, to provide entrepreneurs the safety net they need to pursue their path. For me, that’s been the Air Force—they pay for healthcare and some retirement.
My first venture, in my hometown of Lawrence, Kansas, tackled the lack of broadband services in low-income communities. At that time, people did not identify broadband as a utility and—in misunderstanding broadband as a luxury—questioned why we wanted to build that service for low-income residents. But we made the case that broadband is essential, just like water. Now there is great public support for broadband access and the government is investing hundreds of millions of dollars in its expansion. We’re proud to have been early to that fight, and even more so since the company still serves families today. However, there were regulatory impediments to building a competitive business in that vertical.
From there, I still wanted to facilitate the growth of other entrepreneurs. We thought if we gave folks access to things like low-cost work space, 3D printers, and a data center, those resources could build community and support founders. So we opened a makerspace, and we wanted it to be voice-enabled. And that is how Mycroft was born. This was before smart speakers existed. We set out to build a voice assistant—and wanted to create something where the user could feel they were having a real conversation. Now the voice assistant market is the fastest growing tech sector.
Can you tell us more about Mycroft AI? What is the technology you are building and where is the company today?
Mycroft AI is an open-source voice assistant with fully customizable capabilities. It can be installed and run on a variety of devices including desktops, inside cars, or even on our own Mycroft Mark I and II speakers. It has several standard functions like home control, media activity, and more. With our open-source model, the software is continuously learning and can be adapted or changed based on the preference of the user.
Since our initial product development stage, we have joined a few accelerator programs and raised capital. I was able to attract a new business partner and co-founder, Michael Lewis, who has a track record of successful exits and is a strong supporter of open services and technology. We’re now shipping products and have integrated our software into services from a lot of big corporate brands as well.
What privacy and liability concerns can arise from having an open-source software model and how do you address them?
On the user privacy side, we default to keeping nothing from our users, so there is no data stored for others to obtain. Our system asks users if they’d like to donate their data for the purpose of making the tech better (which about five percent of our users do). Alongside that, we’ve built a system that honors the right to be forgotten. Previously donated data can be erased if a user changes their mind later.
As for openness, it's very beneficial from a security and privacy point of view. In the open-source community there’s a saying: ‘all bugs are shallow with enough eyeballs.’ This means that if you have enough people looking at a code base, then any errors and problems become instantly apparent. With a proprietary code base, companies keep it super secret, which means bugs get pushed into production and companies can end up with substantial vulnerabilities. Openness and user agency are critical. If anyone can look at the underlying tech, then you are able to develop better security and tech that can be trusted.
Last year Mycroft AI signed a letter asking Congress to investigate the increasing rate at which the patent office is exercising its discretion to deny inter partes review. Can you tell us why that issue is important to you?
I would love to see the rules for patent review changed at the federal level. The way things are being interpreted right now, it feels as if the federal government has legalized extortion by patent trolls. It’s one thing when you read about large corporations being sued by patent trolls because they have deep pockets and the resources to cover a defense. However, the way abusive litigants target small companies is enormously destructive. Right now, the rules for patent litigation are very asymmetric and put startups at a disadvantage. We recently got hit with a patent troll suit, and while it is rare for a small company like Mycroft to stand up and fight back, we are. That means, in the context of the troll’s original attempt to get $30,000 from us, that we are looking at spending $2 million to defend ourselves.
The inter partes review (IPR) system is really important, but it needs to be reformed. The whole intent of IPR is to give a company targeted by a patent abuser an alternative path to challenge a bad patent. What we are finding now, under the current policy, is that a company that files with the Patent Trial and Appeal Board (PTAB) within 30 days of being sued has their IPR thrown out because of parallel litigation. This is not what Congress intended in passing the law creating IPR, and it creates a series of challenges for startups. When IPR isn’t there to invalidate a bad patent, startups are looking at $2 million in costs. That creates a big disincentive to fight back against frivolous claims. Mycroft is already trying to establish ourselves and compete with Google and Amazon. Now with the patent troll on our back, it’s that much harder.
Are there any policy issues you think the federal government should focus on to advance AI tech?
Yes, low-quality patents are a big piece of that. Thanks to a Supreme Court ruling, it is already more difficult to get software patents that open doors to abuse, because it should not be enough to patent an idea being performed on a computer.
I think there needs to be a process for third parties to challenge patents before they’re issued, at no cost to the challenger. Democratizing the patent examiner’s job would go a long way towards improving innovation. There is also too much litigation in the U.S. being pursued by non-practicing entities (NPEs). This litigation is making us less competitive on the global stage. Fee shifting or encouraging courts to properly apply Rule 11 would help reduce frivolous litigation. Many times an asserting entity will make a bunch of frivolous or bogus accusations and accuse a company of infringement without doing basic diligence. In our case, for example, we are being accused of infringing patents that have to do with mobile devices, but we don’t ship mobile devices and we don’t support mobile software, so obviously we don’t infringe.
Beyond that, for AI algorithms to be built in the U.S., there needs to be clarity around data ownership and data access. There are no clear brackets around what should and should not be done. This uncertainty ends up creating room for people engaged in nefarious activities to push forward. At the end of the day, that will inevitably damage the reputation of the industry as a whole. We also need legislation to clarify data ownership. They say data is the new oil, and people should have rights around how their data is obtained and used. Without clarification about ownership, it creates barriers to entry for smaller players.
Do you have other suggestions on how policymakers can better support startup founders?
Some states, like Missouri, have a bad faith patent assertion statute that allows companies to sue over certain abusive litigation pursued by NPEs. The downside is that, often, the NPE is an empty shell company with no assets, so it doesn’t do any good to sue them. In Hawaii, as an economic development measure, local officials, lawmakers, and business owners are looking at allowing companies that have a significant presence in the state to be able to go after the NPEs, but also anyone that held a financial stake in that bad faith patent assertion (lawyers, investors, lenders, etc.). More companies will create a significant presence in Hawaii if they can recover their litigation cost. This would also encourage companies to fight back in the first place if they thought it might be possible to recover damages.
Where do you see Mycroft AI two years from now? Five years?
In two years, I expect we will be shipping substantial quantities of smart speakers and have other companies shipping products with Mycroft’s software on it. We are also working on creating self-reinforcing data loops. Today, by default, we keep nothing about our customers unless they want to donate their data. The next step in that process is making donated data available to the community for classification and organization. Essentially, we want to democratize the process of classifying data. In the near term, this can improve wake word spotting and intent parsing, but ultimately will improve the ability of the machine to learn how to deal with full conversations. For example, as community members classify “knock-knock” as attempted humor, the algorithm will learn how to handle knock-knock jokes.
As for five years, we hope to grow the conversational capabilities of our software and have it operate as the user’s agent. Users will be able to have a conversation about the news cycle or book a table at a restaurant. This also differentiates us from competitors. There can be complex incentives at play when, for example, a voice assistant makes a recommendation. We think it’s important to have voice assistants that are really focused on working as agents on the behalf of the customers—and giving recommendations that are in the user’s best interest. So, in five years, I hope Mycroft will allow users to change things in the UX, like voice, wake word, and more—and customers will know that if they ask for a restaurant recommendation, they get the closest options with the best rankings, and don’t just hear about those restaurants that paid to be at the top of the list.
All of the information in this profile was accurate at the date and time of publication.
Engine works to ensure that policymakers look for insight from the startup ecosystem when they are considering programs and legislation that affect entrepreneurs. Together, our voice is louder and more effective. Many of our lawmakers do not have first-hand experience with the country's thriving startup ecosystem, so it’s our job to amplify that perspective. To nominate a person, company, or organization to be featured in our #StartupsEverywhere series, email edward@engine.is.
