Meeting Mycroft: An Open AI Platform You Can Order Around By Voice

Mycroft developer Ryan Sipes, speaking from the show floor of this year's OSCON in Austin, Texas...

(see our video interview here), says that what started out as a weekend project to use voice input and some light AI to locate misplaced tools in a makerspace morphed into a much more ambitious, and successfully crowd-funded, project -- hosted at the Lawrence Center for Entrepreneurship in Lawrence, Kansas -- when he and his fellow developers realized that the state of speech recognition and interfaces to exploit it were in a much more rudimentary state than they initially assumed.

How ambitious? Mycroft bills itself as "an open hardware artificial intelligence platform"; the goal is to allow you to "interact with everything in your house, and interact with all your data, through your voice." That's a familiar aim of late, but mostly from a shortlist of the biggest names in technology.  Apple's Siri is exclusive to (and helps sell) Apple hardware; Google's voice interface likewise sells Android phones and tablets, and helps round out Google's apps-and-interfaces-for-everything approach. Amazon and Microsoft have poured resources into voice recognition systems, too -- Amazon's Echo, running the company's Alexa voice service, is probably the most direct parallel to the Mycroft system that was on display at OSCON, in that it provides a dedicated box loaded with mics and a speaker system for 2-way voice interaction.

The Mycroft system, though, is based on two of the first names in open hardware -- Raspberry Pi and Arduino --  and it's meant to be and stay open; all of its software is released under GPL v3. The initial hardware for Mycroft includes RCA ports, as well as an ethernet jack, 4 USB ports, HDMI, and dozens of addressable LEDs that form Mycroft's "face." That HDMI output might not be immediately useful, but Sipes points out that the the hardware is powerful enough to play Netflix films, or multimedia software like Kodi, and to control them by voice. Unusually for a consumer device, even one aimed at hardware hackers,  Mycroft also includes an accessible ribbon-cable port, for users who'd like to hook up a camera or some other peripheral. Two other "ports" (of a sort) might appeal to just those kind of users, too: if you pop out the plugs emblazoned with the OSI Open Hardware logo, two holes on each side of Mycroft's case facilitate adding it to a robot body or other mounting system. 

The open-source difference in Mycroft isn't just in the hacker-friendly hardware. The real star of the show is the software (Despite the hardware on offer, "We're a software company," says Sipes), and that's proudly open as well. The Python-based project is drawing on, and creating, open source back-end tools, but not tied to any particular back-end for interpreting or acting on the voice input it receives. The team has open sourced several tools so far: the Adapt intent parser, text-to-speech engine Mimic (based on a fork of CMU's Flite), and open speech-to-text engine OpenSTT.

The commercial projects named above (Siri, et al) may offer various degrees of privacy or extensibility, but ultimately they all come from "large companies that work really hard to mine your data" and to keep each user in a silo, says Sipes. By contrast, "We're like Switzerland." With Mycroft the speech recognition and speech synthesis tools are swappable, and there's an active dev community adding new voice-activated capabilities ("skills") to the system.

And if you can program Python, your idea could be next.  

Views: 312

Comment

You need to be a member of Codetown to add comments!

Join Codetown

Happy 10th year, JCertif!

Notes

Welcome to Codetown!

Codetown is a social network. It's got blogs, forums, groups, personal pages and more! You might think of Codetown as a funky camper van with lots of compartments for your stuff and a great multimedia system, too! Best of all, Codetown has room for all of your friends.

When you create a profile for yourself you get a personal page automatically. That's where you can be creative and do your own thing. People who want to get to know you will click on your name or picture and…
Continue

Created by Michael Levin Dec 18, 2008 at 6:56pm. Last updated by Michael Levin May 4, 2018.

Looking for Jobs or Staff?

Check out the Codetown Jobs group.

 

Enjoy the site? Support Codetown with your donation.



InfoQ Reading List

AWS CodeBuild Now Supports Mac Builds

Amazon has recently announced that AWS CodeBuild, its managed build service, now supports building applications on macOS. However, due to Apple's licensing requirements, developers must still reserve a dedicated macOS fleet to utilize this new option.

By Renato Losio

Optimizing Continuous Deployment at Uber: Automating Microservices in Large Monorepos

Uber's microservices architecture, consisting of thousands of services, requires a reliable and efficient system for deploying updates, security patches, and new features. To ensure this process is safe and timely, Uber embraced continuous deployment (CD), automating deployments to production. This has been essential for maintaining code quality and minimizing delays in delivering changes.

By Claudio Masolo

Article: InfoQ AI, ML and Data Engineering Trends Report - September 2024

InfoQ editorial staff and friends of InfoQ are discussing the current trends in the domain of AI, ML and Data Engineering as part of the process of creating our annual trends report.

By Srini Penchikala, Mandy Gu, Namee Oberst, Roland Meertens, Anthony Alford, Daniel Dominguez

Podcast: Engineering Leadership: Balancing Autonomy, Growth, and Culture with Michael Gray

In this podcast Shane Hastie, Lead Editor for Culture & Methods spoke to Michael Gray of ClearBank about engineering culture and leadership.

By Michael Gray

Leveraging the Transformer Architecture for Music Recommendation on YouTube

Google has described an approach to use transformer models, which ignited the current generative AI boom, for music recommendation. This approach, which is currently being applied experimentally on YouTube, aims to build a recommender that can understand sequences of user actions when listening to music to better predict user preferences based on their context.

By Sergio De Simone

© 2024   Created by Michael Levin.   Powered by

Badges  |  Report an Issue  |  Terms of Service