Seven years ago, Google’s artificial intelligence technology was barely able to identify cat videos with passable precision. So what can its AI do today when it comes to tackling actual problems—say, famine, flooding, pollution and parent engagement?
Those are among the social issues addressed by 20 non-profit and educational institutions that are getting grants from the Google AI Impact Challenge, announced today at the company’s annual I/O developer conference. They will split a total of $25 million from Google.org, receive mentoring from the company’s AI developers, and participate in Google Developers Launchpad, a six-month accelerator program normally reserved for technology startups.
The competition received 2,602 applications from 119 countries. Among the 20 recipients are universities from Beirut, Uganda and the U.S., along with a handful of nonprofit organizations building education software for teachers, students and parents. They include The Trevor Project, which plans to use use natural-language processing and sentiment analysis in text and chat messages to support LGTBQ youth who may be at risk of depression or suicide.
Another organization, Quill, is receiving a $1.3 million grant to support its efforts in helping students become stronger writers using AI feedback. Founded in 2013, the New York City-based organization offers web-based exercises that asks students to fix and improve sentences. The tasks cover a variety of grammar skills of different complexity, from basic punctuation to run-on sentences and logical connections.
Quill has recently expanded its exercises to focus on the use of conjunctions as a key mechanic to helping students write construct clear sentences. Using conjunctions properly, says Peter Gault, its co-founder and executive director, is a sign that “you’re not just improving your writing, but expressing logical relationships and improving your understanding of ideas.”
During the current school year, Gault says more than 1 million students across roughly 10,000 schools will have tried an exercise. A forthcoming efficacy study, led by a researcher from the University of Nebraska-Lincoln, will be published this summer.
With support from Google.org, the company’s philanthropic arm, Quill hopes to fine-tune its AI to automatically deliver more precise feedback for students as they work through its exercises. Currently, human editors manually grade sentences to create a data set that is used to train the AI system. Gault teases that a new product is in the works, which will cover critical-thinking skills. These efforts will also mean upping its headcount, from its current staff of 11 employees.
Another grant—$1.5 million—goes to TalkingPoints, a San Francisco-based developer of a family engagement platform that facilitates multilingual communication between educators and parents. The tool essentially allows teachers to send messages via text, web or mobile app that can be automatically translated into 25 languages.
Providing multilingual support to connect families and schools is key to serving a shifting demographic of learners, says Heejae Lim, founder and executive director of TalkingPoints. As the U.S. shifts toward becoming a minority-majority country, the population of English language learners has also grown. ELLs made up roughly 10 percent of U.S. public school students in 2015, and that number is expected to rise.
Like Quill, TalkingPoints relies on humans to create a dataset, which is then used to train the AI translation engine. It will lean on Google’s experts to refine that system. The nonprofit will also add the ability to provide third-party resources to users based on the context of a conversation. For example, Lim says, if the conversation is about an upcoming parent-teacher conference, and the mother is unfamiliar with this tradition, the system can deliver links to information to help her prepare for what to expect.
TalkingPoints claims its messaging tools reached 300,000 parents and teachers last year. Lim believes that number will grow to a million users, across about 10,000 schools, by the end of 2019. Ninety percent of those schools receive Title I federal funding (which is earmarked for low-income communities) and two-thirds of its users reside in non-English speaking neighborhoods, she adds.
A study by WestEd, an education research organization, suggests parents have more conversations with their children after using TalkingPoints. Further studies are planned this summer.
An Alternate Path to Growth
To date, Quill and TalkingPoints have received $5 million and $4.8 million, respectively, in grant funding (including this latest). They operate similarly in offering a basic version of their tools to teachers for free, while selling premium plans with additional features for school- and district-wide adoption. Quill currently claims about 200 such paying subscribers; TalkingPoints counts roughly 100.
Neither are currently cashflow positive, and both have competitors in the for-profit sector, some of whom have raised much more in the form of venture capital. For every Quill, there is a NoRedInk; for every TalkingPoints, a ClassDojo or Remind.
Venture capital comes with expectations of financial returns, often multiples of the original investment. Lim has reservations about how realistic that outcome is for her small team of eight, and given the low-income communities they aim to serve. “I see a future where we can be sustainable in 3 to 5 years,” she tells EdSurge. “I don’t see a clear a path for us to be the next unicorn.”
She adds: “I want to be able to make products that are first and foremost impactful, versus thinking about the bottom line. In a lot of the communities we serve, trying to do both would be challenging.” For schools in poorer regions, she says she provides the service at no or low cost—and at a loss.
Gault echoes that sentiment. “Sometimes, for-profits can have an incentive to focus on wealthier districts that can afford to pay more,” he says. And for now, he intends to stick with Quill for the long haul. “Venture-backed companies tend to have a 7-10 year timeframe, after which investors typically push toward an acquisition or exit. We’re looking at a much longer timeframe.”
