Soapbox Labs, XPERI, and National University of Ireland Get €6.9 Million Grant to Build a Privacy-by-Design AI Platform
Ireland’s Disruptive Technologies Innovation Fund (DTIF) announced €65 million in grants on Saturday for 16 different projects including one that is focusing on the intersection of voice, AI, and privacy. A consortium of Soapbox Labs, XPERI, and the National University of Ireland, Galway (NUIG) received more than 10% of the total with a €6.9 million grant for a new privacy-by-design AI project for on-device voice and image recognition.
The Data-center Audio/Visual Intelligence on-Device (DAVID) initiative will focus on developing a “multi-modal AI platform, encompassing both voice and vision, that offers…more natural interactions with smart toys, games, AR/VR, and smart home devices,” according to a Soapbox Labs statement. Soapbox Labs along with XPERI will also invest in the project directly bringing the total initiative funding to €12 million. Patricia Scanlon, CEO, and founder of Soapbox Labs commented:
The smart toy market represents a huge opportunity for SoapBox Labs but to win it we must first deliver our innovative kids’ voice technology as an embedded solution, so that no data goes to the cloud for processing and kids’ privacy is completely protected.
Toys, Kids, and Privacy
Soapbox Labs is known for its speech recognition solutions optimized for children. Since children have different vocal ranges and articulation than adults they sometimes struggle to be understood by voice assistants designed for the general population. Soapbox Labs has already developed voice-interactive solutions for education and another logical segment where voice will impact children is toys. However, anything that involves children and voice recognition also runs up against both consumer and regulatory considerations around privacy. Sending data to the cloud for speech recognition creates a different set of hurdles for making parents comfortable and complying with legal restrictions.
This is where on-device or edge-based processing comes in for voice. Soapbox, along with XPERI and NUIG will collaborate on building solutions for toys where speech recognition and even visual identification will occur completely on the device with no need for cloud connectivity. In this case, “privacy-by-design” has multiple elements, but the core privacy protection is that no one outside the home will have access to private personal data about children. In the U.S., this should help with Child Online Privacy Protection Act (COPPA) compliance. The popular consumer voice assistants all today transmit voice data that comes from children to the cloud where additional steps need to be taken to comply with COPPA and other laws.
XPERI’s place in the consortium is particularly important because the initiative will need chipset optimization and involve integration into new or existing devices. These are all assets XPERI brings to the table.
Expanding Soapbox’s Market
Beyond the on-device processing angle, the announcement also signals a strategic move by Soapbox into multimodal toys and games. DAVID includes both government grant funding and direct investment by the participating companies. Soapbox is “investing millions” of Euros into this initiative according to a company spokesperson.
Hexa Research estimates that the global smart toy market will be nearly $25 billion by 2025 and will show a 15.5% compound annual growth rate (CAGR) between 2017 – 2025. PullString, which was acquired by Apple in late 2018, was originally known as ToyTalk and was targeting this very market. However, it pivoted out of the toy market in 2016 because of the privacy considerations. “When working with children, you are beholden to some pretty strict laws, and ToyTalk as a company had to work diligently to ensure we toe the line,” said Oren Jacob, CEO of Pullstring in an interview with TechCrunch.
Mattel collaborated with ToyTalk/Pullstring for its talking Barbie doll that quickly was denounced in 2015 by privacy advocates as “surveillance Barbie” due to its WiFi connectivity. Soapbox Labs can avoid this fate by doing all of the processing on the device without the WiFi connection back to the cloud. This will open up a big market where voice interactivity would appear to be well-suited, children at play, with many of whom know how to talk but don’t yet have reading or well-developed technology usage skills.