The Register Citizen (Torrington, CT)
‘It’s amazing’: Danbury man helps Google test AI technology
DANBURY – The coronavirus crisis has locked down the regular activity Michael Montgomery enjoys most.
“Yes, it has – dramatically,” says Montgomery. “I can’t go to TIP Squad.”
TIP Squad, which stands for Technology and Innovations for People, is the piloting lab at Bethel-based Ability Beyond, where Montgomery and his peers test new devices and teach those served by the nonprofit how to use them.
So it was a godsend when Google enlisted Montgomery to help develop its voice recognition technology for people with impaired speech. Google’s artificial intelligence initiative, known as Project Euphonia, is in the early research phase, but it is already paying off with a renewed sense of purpose for the 34-year old Redding native.
“It’s amazing,” Montgomery said during a Zoom interview from his Danbury group home this week. “Without this, I would be going crazy.”
Montgomery has been at work all year training voice recognition technology installed on a Google Pixel phone how to understand his speech. Last year, he recorded 3,000 phrases and has spent much of this year fixing glitches. What kind of glitches?
“It didn’t know Laurie’s name,” Montgomery said.
That won’t do. As the Senior Leader of Empowering Technology at Ability Beyond, Laurie Dale is one of Montgomery’s heroes, along with his mom and dad.
“On the app you can go in and edit things and make it right, so it learns,” Dale said.
For Montgomery, who has a rare disorder known as Lesch Nyhan Syndrome, trying to speak so that others understand him has been a battle.
“Definitely,” he says, adding that when he can’t make someone understand his speech, he takes it out on himself.
“I get very annoyed,” he said with a knowing smile. “I get impatient.”
But that’s happening less and less as Euphonia learns Montgomery’s “non-standard” speech patterns and renders them as voice commands.
“I can control my T.V., I can control my life,” Montgomery said. “And I can talk about all of that.”
A spokesperson for Google cautioned that the Euphonia project was in the early research phase, and the promising applications of the technology were not imminent.
“At this point we don’t have a particular timetable to share, but we’re working every day on how to approach and utilize voice samples and speech patterns of people with neurological conditions,” said spokesperson Zoe Ortiz. “This is something we are incredibly committed to.”
Ortiz noted that one of the early applications of the technology is creating real-time text transcripts to help people with difficulty speaking to make clear connections.
“If there is a tester who needs to go to the grocery store and can use this (real-time transcription feature) to help that grocer to understand what they need, that is incredibly helpful,” Ortiz said.
“It is only going to further their independence.”
At the end of the day, Google’s goal is nothing short of a speech recognition system that understands everyone’s voice.
For that to happen, the Big Tech giant will need to add to its database of non-standard voices.
To date, Google has collected about 1,000 hours of non-standard speech from 1,000 testers like Montgomery.
For Montgomery’s work, he receives a stipend and the assurance that he is doing his part to change the world with his voice.
“Yes,” Montgomery agreed with a nod. “Definitely.”