Mozilla is Embracing Crowdsourcing to Enhance its Voice Recognition SoftwareBy Michael Contreras
Last week, Mozilla, best-known for its web browser Firefox, launched a project to catch up to the strides in artificial intelligence (AI) made by the world’s largest tech companies. In particular, this initiative, called Common Voice, is a crowdsourcing campaign designed to collect speech samples to build a robust voice recognition software to compete with Siri, Alexa, and Google Home. The robustness of AI depends on large amounts of available training data. As a result, companies that have been collecting and storing data for years have a significant advantage in developing AI technologies faster than companies that lack these resources.
Crowdsourcing data to beef up AI algorithms is not unique to Mozilla. About a month ago, another tech company, CrowdFlower, made headlines for its human-in-the-loop crowdsourcing efforts. This reliance on human input reinforces the crowd element as a necessary piece to solving the puzzle of developing useful AI. The pattern further highlights the symbiotic relationship between crowdsourcing and AI. AI depends on crowdsourcing to train and refine algorithms, and crowdsourcing relies on AI to showcase it as a mechanism for gathering critical data.
The Common Voice project involves asking volunteers to submit samples of their voice, which will then be fed into Mozilla’s computers to build a voice recognition system. For Mozilla, the challenge is to accumulate voice recognition data in a way that “the people who are creating the data also benefit” according to the company’s VP of emerging technology. In turn, Mozilla hopes Common Voice will provide individuals with a more personalized AI experience than the current voice recognition services on the market. To learn more about the Common Voice project and to donate a sample of your voice, click here.