Hands-Free AI for Busy Workers and On-Device AI-Powered Transcription
Speaking Naturally is a by-product of our team’s in-depth research into the impact of AI on speech and text, covering conversational AI, chatbots, natural language processing (NLP), speech analytics and everything in between.
This week we see an AI which whispers into the ears of busy workers who do not have a spare hand and much more, including:
- 💷New London conversational AI startup Series A announcement
- 📱Deep Learning Transcription compact enough to fit on your phone
- 👨💻Apple acquires a team of ex-Googlers to make Siri smarter
Learn from the Pros
AI on your phone for offline transcription
Deep learning has made significant accuracy improvements to speech recognition in the past few years but it has been far too large to fit on a mobile device.
For the first time, Google has created an end-to-end, all-neural speech recogniser compact enough to fit on a phone, so that real-time speech transcription is always available (even offline). As you speak, your phone can now output words character-by-character, just as if someone was typing out what you say.
Read more
Apple acquire more to boost Siri’s IQ
Siri was an acquisition and initially put Apple at the head in the voice assistant race. However, Apple then fell behind Alexa and Google Assistant in many ways. Siri is particularly lacking in its ability to answer general information and commerce-related queries, a recent study found.
Acquiring Laserlike could be a way for it to catch up. Laserlike was founded in 2015 by three former Googlers, all with backgrounds in search, and have the pedigree to improve Siri’s search and personalisation capabilities.
Read more
Making an Impact
Hands-free AI for workers in the field
Manual tasks, like carrying out an aircraft inspection, tidying a hotel room, and fixing a car demand the full use of a persons hands, meaning it is more difficult to interact with any connected devices at the same time. This is very limiting, especially since that device may be able to directly aid the manual task at hand (by providing instructions, directions or allowing them to ask questions if needed).
Whispr is a new voice guidance platform that “whispers” instructions and expertise into workers’ ears as they do their jobs and allows them to ask questions back to the device, hands-free.
Read more
Numbers that Matter
London-based NLP startup PolyAI: Series A funding
“Attempts to bring modern AI to customer support have been largely unsuccessful” says Nikola Mrkšić, CEO of PolyAI. PolyAI’s CTO, Shawn Wen, explains: “Our AI agents learn by listening to humans”. PolyAI is a London startup focusing on conversational AI. Founded in 2017 with a core team of scientists and engineers from Cambridge’s Dialog Systems Group, PolyAI are happy to announce their multi-million-dollar Series-A funding.
Read more
Under the Hood
Privacy-friendly datasets with TensorFlow Privacy
Many parts of the world have legislation with stringent requirements for privacy. Combine this with the hunger of machine learning capabilities for data, and you have a complex task at hand.
Google’s TensorFlow Privacy is a promising, free tool in this space.
Read more
