Artificial Intelligence - Episode 1
Host: Bradley Metrock (CEO, Score Publishing)
Guest: Joshua Montgomery (CEO, Mycroft AI)
Duration: 28 minutes, 9 seconds
Podcast Links:
In this debut episode, host Bradley Metrock (CEO, Score Publishing, and host of VoiceFirst.FM podcasts This Week In Voice and The VoiceFirst Roundtable) interviews Joshua Montgomery, CEO of Mycroft AI. The fascinating discussion spans the origin of the company, how the company got its name, Montgomery's views on whether AI is an existential threat to humanity, and the need for an ecosystem of smaller companies to check the power of larger ones.
Transcript:
[intro music]
Bradley Metrock: [00:00:13] Hi and welcome to the very first episode of AI, a new podcast from VoiceFirst.FM. The intent of this podcast is to examine all issues and all angles on the emerging sector of artificial intelligence. In our very first guest, we are thrilled to have Joshua Montgomery, CEO and co-founder of Mycroft AI. Joshua, say hello!
Joshua Montgomery: [00:00:42] Hello! It's great to meet everyone. I'm glad to be here.
Bradley Metrock: [00:00:45] Thank you very much for setting this time aside. What you guys are doing is very very interesting. Since this is the very first episode of this new show, I want to start with just very simply the question of "what is artificial intelligence.
Joshua Montgomery: [00:01:02] Artificial intelligence is technology that uses machines and machines learning to create software and experiences that improve over time based on user feedback and data. That's a long explanation but that's the current definition, I would argue, of artificial intelligence. It uses a computer system and it gets better over time by using data and user feedback.
Bradley Metrock: [00:01:31] Interesting - yeah, I didn't think that was that long. That's extremely succinct. My next question of course is what is Mycroft AI? What are you guys trying to do?
Joshua Montgomery: [00:01:42] Here at Mycroft, we've built an open source community that is pursuing the goal of building an AI or an intelligent agent that runs anywhere and interacts exactly like a person. Our concept as a company is to build an experience for users that when they speak to the technology, they can't tell whether or not they're speaking to a person or a computer.
Bradley Metrock: [00:02:10] How many users are using your software right now? Do you have a rough estimate for us?
Joshua Montgomery: [00:02:17] Sure. It's about 7500 users are using in the software, in one way or another. Of those, 1431 are developers that are contributing in some way, shape, or form, and that community is growing at around 5 percent week over week.
Bradley Metrock: [00:02:35] That's pretty significant. That's great.
Bradley Metrock: [00:02:39] Where did the name Mycroft come from?
Joshua Montgomery: [00:02:43] So in 1963, Robert Heinlein - who's sometimes called the Dean of Science Fiction - wrote a book called The Moon Is a Harsh Mistress. And the book was full of a bunch of fairly new ideas - you know, from that book really comes the concept of CGI (computer generated images) with the AI in the book taking the appearance of a human by rendering itself as a model in a videophone. From that book comes the concept of cyber-warfare - the idea that you can use computer systems that are connected to IoT, which in the case of the book, is atmosphere systems, lights and municipal services like sewers and electricity, as a form of warfare in taking over somebody's home or taking over their services using a computer system in order to advance a political goal.
Joshua Montgomery: [00:03:44] And then also comes the concept of artificial intelligence as an amalgamation of skills or abilities, meaning that the the AI in the book comes into existence because it has a long list of skills and a long list of equipment that's connected to, and over time as more skills and equipment are added, the technology appears to the protagonist in the book as a person or as an artificial entity.
Joshua Montgomery: [00:04:15] And then, finally, the AI in that book...it's one of the few cases, I think, in early science fiction or in modern science fiction where instead of, you know, the Mary Shelley Frankenstein image of the technology coming back to bite its creator, that the AI really plays the role of a protagonist as a positive force within the novel. And then in that novel, the AI is named Mycroft, or Mike. So we named our technology after Mycroft because we wanted to really honor Robert Heinlein's contributions to the concepts that really underpin our approach to artificial intelligence.
Bradley Metrock: [00:04:58] Fascinating. So I can tell you never explained that before.
Joshua Montgomery: [00:05:03] Yeah - we have a whole blog post. Actually, if you look at our...we just revised our logos a couple of months ago. Our new logo is actually...shows a circle in the center, with half of it dark and half of light, representing the earth, with a dot over the top of it, which represents the moon. And in a lot of the animations, we're circling the dot around the globe. It really encompasses the story of our name, and the ideas that we're pursuing.
Bradley Metrock: [00:05:32] Very interesting. So your company started with a Kickstarter, is that right? A couple of years ago?
Joshua Montgomery: [00:05:40] Yeah. We were building this technology as an open source project, and as part of the project, we were using Raspberry Pis as our development platform. And so everybody working on the project had a Raspberry Pi in their desk. But then, of course, they also had a pair of speakers that were plugged into it, and plugged into a wall, and then a microphone, and then a USB cable, and then a microphone, or a keyboard, and a monitor, and a power strip...and you know this giant Medusa nest of cables. So, you know, we built this inside of maker space, so I went back to the maker space and used the 3D printer to print up a kind of an oversized hockey puck enclosure so that we could shove all this stuff into an enclosure. And, you know, we looked at that, and scratched our heads, and said "I wonder if anybody wants to buy it?" And this, in many ways, predates the public awareness of Echo - like, at the time, we weren't aware that Echo even existed. And so we took it to Kickstarter and ended up doing quite well - running the largest Kickstarter ever to come out of our state.
Bradley Metrock: [00:06:46] I saw that! Yeah, extremely impressive. Are the economics of Mycroft AI just that it's community funded, because I saw you can still fund it as a community member through your site? Or where does the money come from, for what you're doing?
Joshua Montgomery: [00:07:04] The initial development is being backed by the venture community.
Bradley Metrock: [00:07:09] OK.
Joshua Montgomery: [00:07:09] We've we've got backing from Social Starts, here in Silicon Valley...from Deep Space Ventures in Dallas...from the crowdfunded VC index fund, down in Los Angeles, as well as a host of angels that are interested in the technology. In some cases, so they can bring it into their own companies...in other cases, there's a commercial opportunity. So that's what's really funding the development and the support that we're providing to the community. The long term...our role in this new AI world is to allow everybody who isn't Apple and Amazon and Google and Baidu and Microsoft to deploy intelligent agents within their products and services using an open source stack and we provide enterprise software to allow them to do that at scale. And so our concept is, you know, if you're a single developer, or a single household or family, and you want to deploy the technology, you can use our speech-to-text back-end, and our API aggregation, and just create an account and go. It will give you weather data, gives you access to Wolfram and Wikipedia and a host of other data sources. And that doesn't cost you anything.
Joshua Montgomery: [00:08:33] But if you're a big company - let's say a major retailer based in Bentonville - we obviously would pass across a bill every month for the tools that are required to deploy it across hundreds of thousands or millions of users.
Bradley Metrock: [00:08:51] So on This Week In Voice this week, we were discussing Amazon's new SDK that they just released, which I'm sure you probably saw just in the last 24 hours or so. Do you and Mycroft AI ... do you view Amazon as a competitor? Or do you view them as just another participant in this massive space? How do you look at what they're doing specifically?
Joshua Montgomery: [00:09:20] I think the reality of Amazon's position within the market is that it doesn't matter what business you're in: Amazon is a competitor. This is a company that started off selling books, and has now launched a technology company that is doing space launch services, right? I mean it's a long journey from I'm selling you science fiction novels, to I'm selling you rocket ships, right?
Bradley Metrock: [00:09:53] To I'm writing the science fiction novel...
Joshua Montgomery: [00:09:56] Yeah, exactly. And so, you know, my argument would be not only do we compete with Amazon, but if you're in business, you compete with Amazon, in some way, shape, or form. And as a result, you know, leaders within industry that are capable of strategic thinking - and there's a lot of them that aren't, but there are a lot that are - would be very wise to think twice about sending every consumer interaction that's being conducted on your product or service through our friends at Amazon.
Joshua Montgomery: [00:10:34] I know that on the first day of the deployment of the Internet, the publishing industry saw it as a better way to distribute their news. But by partnering with, you know, large Silicon Valley companies, and by making their content easy to search, and giving their content away for free, the local newspaper industry for example has decimated itself. So you'd be very wise to think twice about fully embracing the Silicon Valley giants. We view Amazon's role in this market as really a leader that's building fantastic technology that a lot of people are adopting, and we think that they're going to be very very successful. But there's a whole host of companies out there, for whatever strategic reason, want to be able to deploy voice technologies on their premises, within their security perimeter, in ways that allow them to be independent of the Silicon Valley giants - to have data independence to maintain the security of their information and to keep their customer relations private. And for those customers, you know, we are able to provide a solution that does that, which really makes us unique. All of the other technologies in the space send all of the customer data, and all the queries, to an online service, where when you read the privacy policies and the privacy statements, there are significant challenges surrounding both consumer privacy, and then, of course, competitive and corporate secrets.
Bradley Metrock: [00:12:15] Interesting. Do you own an Echo at your house?
Joshua Montgomery: [00:12:20] Yeah, I mean we do our homework, so we have voice assistants from all over the world - I'm actually looking at a box that contains a Ling Long Ding Dong from JD.com in China. We evaluate all these systems, and in some cases, we learn from them, and in some cases, they're potential partners. We're very careful to keep up on progress in the marketplace.
Bradley Metrock: [00:12:53] So, recently - and I'm sure you saw this - Elon Musk and Mark Zuckerberg sort of got into it publicly over the subject of artificial intelligence. Elon Musk saying "I'm afraid of it." Mark Zuckerberg saying "you're being dramatic." Of course, I'm paraphrasing. And Elon Musk then coming back saying "you don't have any idea what you're talking about." I am intrigued with your background specifically. You're a successful entrepreneur, but you've also been in the military eight years, in the National Guard if I read correctly, and I'm sort of interested in your perspective. Should we be afraid of artificial intelligence? And if so, how afraid should we be?
Joshua Montgomery: [00:13:42] So I'll start with the disagreement between Elon and Mark. You know, media in general thrives on controversy, and in that case, my read on the situation is that the various media outlets involved are trying to gin up a controversy that really doesn't exist.
Bradley Metrock: [00:14:02] Sure.
Joshua Montgomery: [00:14:03] And so but in terms of whether or not artificial intelligence is an existential threat, I would argue that artificial intelligence in the hands of the few is an existential threat to...certainly to the economy, and through the economy, you know, I would say yeah, it's an existential threat to the safety and security of everybody here on the planet. Let me kind of flesh that statement out a little bit.
Bradley Metrock: [00:14:35] Sure.
Joshua Montgomery: [00:14:36] You know, as we move forward, these artificial intelligence agents that we're building are going to be able to, and in some cases are already able to, significantly change the way that humans work. And so, as a great example, we come up through the Sprint accelerator and through our friends at Sprint, we learned that right around 90 percent of their call volume in their call centers consists of 25 questions. And to build an artificial intelligence that's capable of, in a very natural way, answering 25 questions, borders on trivial. And Sprint employs something like twenty-seven thousand call center workers, to meet that call demand. And so if you deploy an artificial intelligent agent into Sprint, for example, you run the risk of putting, you know, tens of thousands of people out of work, and replacing them with a simple machine learning stack.
Joshua Montgomery: [00:15:36] The danger becomes if the people who own that technology - if it's four guys, or five guys, or 50 guys in Silicon Valley - and none of the ownership of the improvements to technology flow through to the people whose jobs have been replaced. We're on the border, we're on the cusp, of a future where artificial intelligence can really make repetitive work and drudgery obsolete. That pretty much any task that's repetitive can be performed by a robot, or by an artificial intelligence. And in a world where anybody can deploy 100 intelligent agents to go solve a repetitive task, you really create the potential for a utopia. I mean, how awesome would it be for you...you know, at your podcasts, to be able to have 100 agents to help you to find stories and to help you to write stories and conduct interviews?
Bradley Metrock: [00:16:39] Oh, it'd be beautiful!
Joshua Montgomery: [00:16:41] It'd be beautiful. And it would allow you to create so much more. But if all of those agents are owned by 5 guys, then all of the wealth that that creates flows to those five people. It's our view that the real danger with artificial intelligence is that the ownership of the technology is concentrated, not so much that the technology itself is a threat. And it's one of the reasons that we're focused on, and I know Elon is also focused on, opening these technologies and making them available to everybody, because it really does mean that people can work less at repetitive tasks that are drudgerous, and work more on creative tasks and communicative tasks. And depending on how we handle that transition, we really have the potential to make everybody's lives much higher quality.
Joshua Montgomery: [00:17:34] And I'll take this back to the call center example. You know, in that call center example, can you imagine sitting in a call center and answering the same 25 questions all day, every day, forty hours a week? I mean, what a terrible, terrible job. But it's that 1 out of 10 questions that comes through where you can really add value for the customer, where you're able to build a relationship with the customer, where you're able to help that customer solve a problem that the automated systems weren't capable of. That's the question that really, for me at least, if I'm working in a call center, adds value to my day, and makes me feel as though I'm contributing to the mission of the company I work for.
Joshua Montgomery: [00:18:13] And so, you know, that's what artificial intelligence can create: a world where everybody who's going to work is fulfilling...living to their fullest, as a human being, and being creative and collaborative and not doing the same thing over and over and over again. And so, I'm very hopeful about the future of AI. But I certainly see that there is danger, and the danger is that Facebook, Apple, Google, Amazon, Microsoft, Baidu, and a handful of other companies own all of this technology outright, and everyone else is left...you know, as a consumer, and not a creator.
Bradley Metrock: [00:18:58] That's fascinating. And I completely agree with you - I think the reason why voice technology has advanced to this point, where it is right now, which is literally on the precipice, probably over the precipice of public acceptance, is that the tech juggernaut most associated with trust and customer service, which is Amazon, has been the one leading the way. If it had been...you know, I'm not sure even Google could have done it, but you know, if it had been Microsoft or Apple, I think it would have taken far more marketing dollars, far more effort than what Amazon has been able to do in leading the marketplace. It's fascinating.
Bradley Metrock: [00:19:46] I agree with you: I think disseminating the technology is a good check against abuse, and you articulated that really well.
Bradley Metrock: [00:19:58] So paint a picture for me, Joshua: five years from now, if I look in a kitchen in a median household in the United States, am I seeing zero voice assistants - zero pieces of hardware that are voice driven - am I seeing one? Or am I seeing all of them?
Joshua Montgomery: [00:20:28] As to whether or not there are microphones on every device, I really can't answer the question. There very well might be. I think that what you do see is at least one, and possibly you see those devices working together in an intelligent way. I think that it may continue to be a complex situation where, for example, let's take your kitchen as an example. As long as you have an Apple refrigerator, and an Apple blender, and an Apple HomePod, and an Apple phone, and an iMac, the experience will be very tightly integrated and Siri will help you figure out whether or not you need to order eggs. I think that a vast majority of homes will have some kind of a mixed environment, where they will have a voice assistant of some kind, and that voice assistant will be able to access services in their various different appliances and access their various different screens. Right now, there's a big effort being made by all of the major ecosystems to build these walled gardens where their voice assistant...yes, it can access Philip Hue, and yes it can access WeMo, but really it's designed in, for example, Google's case, to work with their Pixel line of phones, and to work with their Google Home device, and to work with their Chromecast on the screen, and really doesn't play nicely with the other ecosystems. Certainly that is the case at Amazon with the Fire Stick and the tablets and the Alexa services.
Joshua Montgomery: [00:22:12] You know, our hope would be that there's room in there for a neutral party that doesn't really...unlike our friends in Seattle, I'm not trying to sell you retail products. Unlike our friends out in Mountain View, I'm not trying to provide a discovery service so that you can find other Web sites and I can sell you advertising. Unlike our friends at Cupertino, I'm not trying to sell you beautiful, highly-priced electronics. The concept would be that I'm trying to provide you with the best voice experience possible, and work across those ecosystems. As to whether or not those vendors will allow that and facilitate that, I don't think that they're going to facilitate it. They may allow it, and consumers may demand it.
Bradley Metrock: [00:23:02] Well I certainly think companies and other organizations might demand it. To me that's quite possibly your bread and butter. You go into somebody's office and paint the picture that you described earlier in this podcast...that you don't want to be giving your information and every single query that every single employee every single day has to say, over to Amazon or Google or whoever, you would get some business out of that. To me, that's extremely logical. But on an individual basis, it could make sense too. I think what you're doing is brilliant.
Joshua Montgomery: [00:23:37] Well thank you very much. We're really excited. And by the end of the year, in English at least, we'll have the technology stack working to the point where you can run it very reliably, completely independent of the internet. So being able to run the entire experience, including STT on site. Now, the speech-to-text component takes a little bit more processing power than is available in a Raspberry Pi, clearly. But with maybe a separate STT device, that's got a little bit more juice in it, you'll be able to run independent of the Internet. And I know that there's a really broad swath of both companies and people that really want to maintain data independence. And the hope would be that...you know, the place where we would really like to be in five years is in a situation where the open alternative, whether it's us or maybe somebody else builds an open stack that successful, and we we wish those folks luck, provides an agent that acts as an agent on behalf of the user, and not an agent that has an ulterior motive on behalf of some service provider. So the example I always use is if you're using an agent that represents a service provider, and the phone rings at 10 PM, and your agent answers, the call I want to go through is the call coming from my mom saying my dad's not doing well. The call the company wants to go through is the call we're the telemarketer paid him $5 to get through my filter. Right? And the question becomes is in a world where your intelligence agent starts managing a lot more of your life - whether it's your email, your text messages, your social media presence, your incoming and outgoing phone calls, your media consumption - who do you want that agent to be accountable to? And, right now, we're in the danger of building a world where there's like five or six companies that those agents are accountable to, and they have ZERO accountability to the public at all. None. And that's a real challenge, and it's the challenge that we're hoping to address.
Bradley Metrock: [00:25:48] Fascinating. Well, I certainly appreciate you setting your time aside, Joshua, to share your insight and expertise with us. Let me ask you a question before we go: are you a Jayhawks fan?
Joshua Montgomery: [00:26:00] Absolutely. I'm a Lawrence native. I grew up, went to Lawrence High School, went to the University of Kansas, started two businesses in Lawrence...I can't speak highly enough of KU.
Bradley Metrock: [00:26:12] Nice. Yeah it's a gorgeous campus. Been there once. You guys going to be good in basketball this year?
Joshua Montgomery: [00:26:19] Always.
Bradley Metrock: [00:26:20] Like the answer to that's going to be no...
Joshua Montgomery: [00:26:21] We shine at basketball. We might be the best at evolution, in Kansas, but we shine at basketball.
Bradley Metrock: [00:26:31] If someone has heard this podcast, and they want to reach out to you, what is the best way to do that? Is there a form on your website, or an email address? What's the best way to reach you?
Joshua Montgomery: [00:26:44] Sure. You can send an e-mail to founders at Mycroft dot AI, and that will reach us. We also have a very active user community in MatterMost, and via our developer forums, and so folks who are interested should absolutely reach out. Today, we're very interested in speaking to developers who are looking to build skills on the platform, or who might have a need to improve the state of speech synthesis or speech recognition. And then, of course, we're looking for corporate partners - for proof-of-concept deployments. We're already working with a number of big brands, including Jaguar Land Rover. But we would love to have the opportunity to work with others.
Bradley Metrock: [00:27:25] Excellent. Joshua, thank you very very much for being so generous with your time.
Bradley Metrock: [00:27:32] For the very first episode of AI, on VoiceFirst.FM, thank you for listening...and until next time.
[exit music]