TOKYO: Picture this. Google's AI system assisted researchers in New Zealand in identifying calls of native birds - Kakariki and Hihi - using acoustic sensors after sifting through 15,000 hours of audio captured in and around Wellington. The Google Brain team, a core group focused on deep learning, used a trained Tensorflow model to label spectrograms and validate results to classify bird songs in real time.
Tensorflow is an open source software for machine learning (ML) developed by the Google Brain team that was launched in 2015.
Since then, Google has been running ML on different data sets - from tracking seacows to diagnosing diabetic retinopathy and other health challenges.
Linne Ha, director of Google Research and Machine Intelligence, shared updates on Google's ambitious Project Unison that's attempting to create text-tospeech (TTS) voices for lowresourced languages. There are 6,000 languages globally and 400 of them have over a million speakers. The effort is to bring more global languages online. Ha talked about how Google crowdsourced voice samples of Bangladeshis speaking Bengali, creating a speech database using machine learning to build a text-to-speech engine. The samples curated by Google were put to vote for the best voice profiles.
At the MadewithAI event in Tokyo recently, Google presented a primer on AIcentric use cases in its portfolio, including Google Photos, Assistant and its smart speaker Home and smartphone Pixel.
Google senior fellow Jeff Dean articulated how AI is shaping our everyday lives. Google aims to make AI more mainstream through its hardware-plus-software portfolio. "We want to help businesses innovate using ML and AI in their own companies. We want to provide internal and external researchers with the right tools to tackle really large challenges in building AI systems in the future," he said.
Dean said the science of ML has taken over the field of AI as machines are exposed to myriad data sources to make predictions. "We can train neural networks for a lot of different kinds of inputs and different kinds of outputs. We can run any picture through the model and it can give predictions. We can take a sequence of audio features and transform it to speech recognition. These are quite complicated functions that systems can learn. That's why people are quite excited about it," he added.
Dean's efforts tie into the company's vision of transitioning itself into an AI-first entity over the next decade where computing becomes universally available. Google is focusing on strengthening its AI talent pool.
It had over 1,000 engineers trained in ML in 2012, but the number has risen sharply to 18,000 today.
As AI is going to be a part of everyday work and life, there are concerns around privacy at a time when a virtual personal assistant is listening and responding to voice commands all the time.
Pravir Gupta, engineering director, Google Assistant, said privacy is of utmost importance and said how it allows users to have the tools and controls for the interactions they have had with Google. "You can go to the activity page on the Assistant, and delete previous activity. Much like search, providing that transparency and control is front and centre," he said.
(The correspondent was in Tokyo at the invitation of Google)