Language is all about repetition. Every word you're reading was created by humans, and then used by other humans, creating and reinforcing context, meaning, the very nature of language. As humans train machines to understand language, they're teaching machines to replicate human bias.
On Thursday, Google announced that its Home smart hub device can now recognize and identify up to six different users by the sound of their voice. It's an inevitable—but crucial—step in the development of smart home virtual assistants. The new skill means that different people in a household will be able to ask the Google Assistant questions about what's on their calendar, or what their commute looks like, and the Home device will know who is speaking to it and give tailored responses. It'll make it a more streamlined experience for families sharing a smart home speaker hub.
The future of bots is sitting in thousands of documents folders, waiting to be born. At least, that's the premise of Albert, a bot and bot-creation tool from NoHold, which released a pro version on Monday. The premise behind Albert is straightforward: upload a document, and then ask the Albert-generated bot to answer questions with information based on that document. Albert is a product of the modern era of chatbots, but Albert's origins are, by tech standards, positively ancient: the key work dates back to a patent filed in 1999.
Today's artificial intelligence is certainly formidable. It can beat world champions at intricate games like chess and Go, or dominate at Jeopardy!. It can interpret heaps of data for us, guide driverless cars, respond to spoken commands, and track down the answers to your internet search queries.
Say you're on the phone with a company and the automated virtual assistant needs a few seconds to “look up” your information. And then you hear it. The sound is unmistakable. It's familiar. It's the clickity-clack of a keyboard. You know it's just a sound effect, but unlike hold music or a stream of company information, it's not annoying. In fact, it's kind of comforting.
Speech recognition software isn't perfect, but it is a little closer to human this week, as a Microsoft Artificial Intelligence and Research team reached a major milestone in speech-to-text development: The system reached a historically low word error rate of 5.9 percent, equal to the accuracy of a professional (human) transcriptionist. The system can discern words as clearly and accurately as two people having a conversation might understand one another.