According to a 2017 University of Washington report, there are hundreds of millions of smart-home devices in more than 40 million U.S. homes. This number is expected to double by 2021.
Amazon Echo, Google Home and other devices that have Alexa and Google Assistant built in, have proven to be some of the world’s most promising new technologies. These AI-enabled assistants seem capable of doing everything, from turning on lights to answering simple and even complex questions. “OK Google” and “Alexa” have become common household phrases, as these smart connected speakers always have their microphones on, yet don’t respond until their “wake words” are mentioned.
While useful, AI-powered assistants are also rather creepy. For instance, after users say the “wake word,” Alexa and Google Assistant start recording conversations, sending that data away to the cloud. Since the technology behind the Amazon Echo and Google Home is powered by AI, it stores command history to help make the device “smarter” to better respond to future commands. However, interactions stored in the cloud are prone to hacking, putting users’ personal data at risk.
The way AI-powered voice assistants function struck a nerve for Dr. Rand Hindi, a recent Forbes 30 Under 30 entrepreneur. Hindi has been concerned about data privacy for a number of years. He started coding at the early age of 10 and founded his first tech company when he was just fourteen years-old.
“The first company I created was a social network. I remember thinking how wrong it was that I could access the network’s database at any time and see the messages users were sending to each other. The idea of data privacy has been on my mind ever since. It also became clear that privacy was going to be one of the main challenges facing the rise of artificial intelligence.”
According to Hindi, leading companies like Google and Amazon are misusing users’ personal data collected from AI-based voice assistants like Alexa, Siri and Google Assistant.
Many companies claim to protect users’ personal data via “privacy trust,” an idea where users can send a company data and the company is trusted to delete that data after the fact. However, no guarantees are made that the data will be deleted. Also, voice assistants are really just a microphone in your home connected to the Internet. It is crazy to think that people are wire tapping their own homes by using this technology, Hindi told me.
“Privacy By Design” With Blockchain
To ensure that users’ data is kept private, Hindi founded Snips, a leading AI company based in Paris and New York that is focused on bringing privacy to AI-based voice technologies. Since 2013, Snips has been building voice technology to guarantee “privacy by design.”
Snips AIR is an example of the privacy by design concept, as it’s an AI voice platform that utilizes blockchain technology to ensure that user data never gets sent to the cloud. Snips AIR, which will be available for consumers at the end of 2019, is a decentralized alternative to Siri, Alexa and Google Home and is fully compliant with the General Data Protection Regulation (GDPR).
Rather than storing user data in the cloud, Snips AIR processes all data on-device, ensuring that personal information remains within the walls of connected homes. And while no one has access to user data (not even the Snips developers), the AI behind the Snips AIR voice assistant can still be improved and trained over time by combining “federated learning” with blockchain technology and modern cryptography.
In other words, instead of users’ data being sent to the cloud for the AI to get smarter, the Snips ecosystem pays users in tokens to contribute their encrypted data to its blockchain network. The processed data is then aggregated by application developers, who now have new machine learning models that contain the encrypted user data. Since training the AI is done on the blockchain with encrypted federated learning, developers can take advantage of user data and add new applications to the Snips AIR platform. The users themselves never have to expose their personal information and compromise their privacy.
The Snips platform (released in June 2017) consists of more than 14,000 developers who have created more than 24,000 voice assistants. According to Hindi, the developer ecosystem is focused on building use cases that future consumers will be able to use. Original Equipment Manufacturer’s (OEMs) can then integrate the Snips assistant into their products and benefit from the entire Snips voice ecosystem, while protecting user privacy every step of the way.
Other blockchain based platforms are also trying to solve this problem. For example, LangNet, a decentralized ecosystem for language AI, recently teamed with Dopamine.ai, the decentralized global marketplace for data and AI consumers and providers. LangNet and Dopamine share the vision of creating decentralized data ownership, while also focusing on monetization opportunities that are provided by the various data providers.
LangNet aggregates a rich and vast voice repository, while Dopamine provides Decentralized Collaborative Processing Networks (DCPNs), in which data and AI providers are incentivized to supply various services that consumers can pipeline for specific needs. LangNet allows anyone to contribute language data in exchange for tokens. Collected data includes voice recordings of users reading a sentence to train speech-to-text systems and personal commands for a voice assistant. Data is owned by the contributor and licensed out to developers in exchange for tokens.
Through this partnership, LangNet will be able to easily integrate with the Dopamine’s platform as the key voice data provider. On the Dopamine’s platform, the various enterprises and developers in the market will have the opportunity to access the LangNet network, which has been designed with a growing source of the language resources. Various campaigns will also run to help with crowdsourcing the voice data sets, along with the models that they will need to assist in building the new application with the use of different languages and also for the new use cases.
The best security measure is zero data. If I have no data on my users, then there is nothing to hack. In these ecosystems, blockchain is not a technology that makes data private, but rather it’s a technology that makes this all work by incentivizing users to contribute and process data in an honest manner. Blockchain is really the glue between all the pieces that have been around for years. The reason I got into blockchain was not because of cryptocurrency, but really because of the idea that we can create self governing communities in a way beneficial for themselves, Hindi said.