Every few years, Silicon Valley becomes obsessed with a new development in the tech space. Some, like the 3D web, fizzle out and leave nothing more than a footnote in the history of the tech industry. Others, like the push towards cloud services, have become a central part of many users' (and developers') lives.

Silicon Valley's latest craze is voice assistants. Google, Apple, Amazon, and a few others are all jostling to make the most capable voice assistant on the market, hoping to create an artificial intelligence that will do everything for you on your phone or computer. This year's CES has seen all kinds of devices powered by either Google's or Amazon's AIs – home automation kits, smart speakers, cars, and more.

As is the case with most proprietary services, these AIs have one big caveat: They are opaque, and they collect a lot of information about you that gets sent to someone else's opaque servers. Luckily, the open-source community is already at work on several voice assistants that are transparent and respectful of your privacy – and that strive to match their commercial rivals in functionality.

Mycroft

Mycroft [1] is the most developed and most popular open source AI assistant. It is created by Mycroft AI, a 19-man operation set up by entrepreneur Joshua Montgomery in Kansas [2]; despite its small size, Mycroft AI has the ambitious plan of creating a virtual assistant that will be able to compete with the devices developed by the giants of Silicon Valley (Figure 1). The company has announced plans to "demonstrate something approximate to a human" by 2020 [3]. The open source development model is an integral part of this plan.

Figure 1: Joshua Montgomery, Mycroft AI's founding father and CEO.

The clearest sign of Mycroft's intent to take on Google and Amazon is their devices. The company has released one smart speaker, the Mark I, for developers. Funded via Kickstarter in 2015 [4], the device resembles an adorable cross between ET and Eve from the movie WALL-E, with a white ovoid shell and two big eyes and a mouth made with LEDs at the front of the device (Figure 2). The company has started another Kickstarter campaign for their Mark II, which resembles a plastic and cloth egg with a screen on the front [5].

Figure 2: The Mycroft Mark I, with its big LED eyes and white plastic shell.

Despite their commitment to developing smart speakers, the company also provides its software for Linux distributions.

Skills

Mycroft organizes tasks into skills. When the user asks a question or gives a command, Mycroft uses an intent parser to identify the reason for the utterance. Once the intent is known, Mycroft checks to see whether it has a skill that addresses the intent. For instance, if you ask Mycroft about the weather, it will parse the request and send the intent data to the weather skill. Mycroft currently installs with 30 core skills when you set up the system [6]. You can think of these skills as the things Mycroft "knows how to do" at the time you install it. (See Table 1 for some examples of Mycroft core skills.)

The beauty of the open source development model is its flexibility and capacity for expansion. Unlike many proprietary digital assistants, Mycroft lets users develop their own skills. The project website includes instructions for creating your own Mycroft skill [7].

Users can submit these custom skills for acceptance and distribution to the Mycroft user community. The website has a long list of community-developed skills that any user can add to their Mycroft configuration. Examples of community-developed skills for Mycroft are shown in Table 2. Some of these skills are currently marked as untested or works in progress, but the examples in Table 2 give an indication of the kinds of tasks Mycroft users are hoping to automate.

This extensive list of core and community skills means you can tailor Mycroft to do what you need it to do. Of course, as you will learn later in this article, some of these skills don't exactly work as advertised at this point in Mycroft's evolution, but if the power of open source development leads to community bug hunting and a vibrant pipeline for newer and better community skill development, the quality and scope of Mycroft's capabilities will progress rapidly.

Installation

The main reason for Mycroft's desktop version is to make it easy for developers to develop for the Mycroft platform [8], but you don't have to be a developer to get it to work. It takes just a few commands to get Mycroft up and running.

git clone https://github.com/MycroftAI/mycroft-core.git
cd mycroft-core
bash dev_setup.sh