The Edtech AI Space Race Has Begun
After years of promises, serious contenders to end "one-size-fits-all" are starting to make headway
Knewton and The Prehistory of Edtech AI
When I first joined Knewton in 2008, fresh out of graduate school, we all had a shared dream of bringing education into the 21st century.
That company was founded (and funded) on a big dream- that technology algorithms could optimize teaching and learning, and bring some of the incredible innovations of the information age to the daily lives of students. Knewton’s pitch was that every student should learn not only at her own pace, but using the materials and methods that are most effective for her- what has been called ‘adaptive learning’, or in some cases ‘personalized learning’.
Knewton’s dream was… premature.
Despite its best efforts, lots of funding and a truly brilliant team, Knewton could not, at the time, gather nearly enough big data or build an accurate enough AI model to become the definitive recommendation engine for students. There were certainly victories; Knewton’s innovative math remediation program at ASU was accelerating learning and reduce attrition for incoming freshman, and Knewton’s GMAT, GRE, LSAT and SAT test prep programs were proving effective, but the real Holy Grail was out of reach.
Moreover, Knewton’s founder, Jose Ferreira, was fond of high-flying rhetoric (such as the fact that Knewton was ‘the robot tutor in the sky’), which led to a deep backlash in the education and business press.
Looking back, Knewton was working on grand scale AI in parallel to the development of powerful plug and play AI tools like AWS Sagemaker (2017) and Google Tensorflow (2015) and Python’s Sci-kit learn package (2010), as well as massive innovations in the artificial intelligence world.
In the last two years, the dream of bringing cutting-edge AI into education to create adaptive teaching and tutoring systems has been revived in a big way, all over the world. In fact, one might argue that Educational AI is edtech’s version of the “Space Race”, in which countries compete to prove their scientific and intellectual dominance.
Let’s run through a few of the contenders who have thrown their hat in the ring for the new Race to Create Educational AI.
Google (US)
Google has making waves in edtech for years now, notably through the dominance of Google Classroom as the go-to open LMS for k-12 schools and the rise of Google Chromebooks. Over the past year, Google has been working on an AI-powered suite of tools that is designed to fit into existing higher education learning systems, which was announced this month on the Google Cloud blog.
According to Google Cloud Head of Education Steven Butschi:
The platform is a suite of applications and APIs that can be integrated into an institution's existing infrastructure. Educators build competency skills graphs that feed the platform, which then uses AI to auto-generate learning activities for students. Learners can easily access these activities through an app—Google Cloud’s customizable interactive tutor solution for educational institutions.
The suite is being put to the test in several pilots, including a high-profile one with Southern New Hampshire University, the innovative ‘megauniversity’ that has been popularizing competency-based education with non-traditional students for years.
The killer app here, if it works, is the potential to plug the Google tools directly into “existing infrastructure”; this could massively accelerate adoption- or at least trial adoption- which would then lead to the collection of big educational data needed to optimize the tech.
Byju’s (India)
India’s Byju’s has become by far the biggest Edtech unicorn globally, with an $21B+ valuation after its most recent Series F round of investments (which includes both Chan Zuckerberg and Google money).
Byju’s has acquired nine strategic edtech companies in 2021 alone, and is clearly building a massive suite of services in preparation for an alleged IPO in 2022, including AI tooling.
An Inside Higher Ed article yesterday cited how Byju’s is moving into the AI space in a big way, and according to the Byju’s Lab announcement:
“The role of online learning is not just to replicate offline classes in digital space but to make it more interactive, engaging, and personalised. By combining the ability of computing, technology, and data, we at BYJU’S Lab, want to explore the power of information and technology to create a more personalised, enhanced, and democratised learning”
While not much has been released about this initiative so far, Byju’s certainly has the funding, the expertise and the ambition to make this dream into a reality; and as one of the most well-known start-ups in any sector in India, they will have access to new talent. They have set a timeline for three years for these tools to be released.
Squirrel AI (China)
Squirrel AI is a Chinese edtech that uses AI technology in the tutoring space; its conceit, like others in this list, is to provide students with a virtual “AI Super Teacher” that can be more effective than either their classroom teacher or a private tutor- to solve, or even surpass, the Bloom’s Two-Sigma problem.
Squirrel boasts Tom Mitchell, the former Dean of the School of Computer Science at Carnegie Mellon University, as its “Chief AI Officer”. Its methodology, as described on its website, seems like a turbocharged version of Knewton’s, creating a massive taxonomy of knowledge components and matched learning content so that students can be served appropriate and helpful material on any question or topic, and a feedback loop can be used to ‘train’ the AI model to become effective.
According to Squirrel’s site:
“The Squirrel Ai Learning Intelligent Adaptive Learning System (IALS)… pioneers the decomposition of nanoscale knowledge components, take junior high school mathematics as an example, 300 knowledge components are dissolved into 30000 fine-grained knowledge components, and each knowledge component is matched with learning content, including text item, animation, slides, short instructional video, etc.”
The Chinese tutoring world has been upended recently by government policy, but its too soon to tell whether the changes will benefit or hurt Squirrel- one can imagine that without an army of non-Chinese teachers at the ready (as VIPKID offered until recently), a Chinese-developed AI super teacher might be just what the doctor ordered. Whether or not Western students (and parents) are open to it is another matter.
Allen Institute for AI (US, Israel)
The Allen Institute for AI, staffed with top AI research scientists from both the US and Israel, has several projects that are designed for education, especially in the realm of Natural Language Processing. The Aristo project, for example, can parse text, make sense of it, answer questions about it, combine pieces of information, and more:
Because the institute is research-focused, it is likely that the project designed here will need to be ‘productized’ or licensed by consumer companies, the same way that SRI developed the voice technology used in Siri ; given Paul Allen’s relationship with Microsoft, one could easily imagine Microsoft using this technology in their growing suite of education tools.
Sana Labs (Sweden)
Sana Labs is a company out of Sweden that promises “personalized, collaborative learning”. The ‘collaborative’ part is intriguing, and indeed, Sana’s product demo emphasizes a Google Docs-like synchronous collaboration tool, with multiple learners working together.
Joel Hellermark, Sana’s CEO, is a highly visible frontman with a somewhat Ferreira-like flair for the dramatic; see Sana’s recent “Frontiers of Learning” video for an overview of their vision of AI-powered Edtech.
Sana got an $18m Series A this year at the beginning of this year, and their emphasis on personalization and collaboration (as well as their gorgeous Scandinavian design) makes them a dark horse contender to be the winner of the AI Edtech Space Race.
Other Global Contenders
Sense Education (Israel) offers the ability to leverage AI to scale human feedback, allowing single professors to provide thousands of students with feedback. It seems to have been used in Georgia Tech’s disruptive $7,000 OMSCS Computer Science Degree.
Atama+ (Japan) received the 2020 Japan e-learning award for its personalized education tool, which so far is used primarily in preparatory schools. Its stylized platform and branding seems designed to appeal directly to students.
Teach to One 360 (US) has been building its research-backed, math-based personalized learning program for over a decade and is designed to incorporate into the US school system.
Classting (South Korea) is a widely popular social platform with over 7,000,000 users in Korean schools, that uses AI to personalize learning.
Update, 12/1: A particularly astute reader suggested a couple of other Edtech AI companies that have thrown their hat into the edtech AI Space Race, both of which certainly deserve mention here. (If there are more, please let me know on Linkedin):
RIIID Labs (South Korea) offers a suite of tools to predict assessment scores from shortened assessments, maximize learner engagement and design personal learning paths . Their RIIID Tutor program has over 2,000,000 users, and the company has raised $175 million to fund its ‘global expansion’, starting with Japan.
Area9 Lyceum (Denmark) has a long track record of educational innovations, and it’s new Rhapsode system claims to be the first ‘multidimensional’ adaptive system (the dimensions are knowledge, skill mastery, metacognition, and, interestingly, character). Area9 already has offerings across the learning spectrum, from K-12 through Higher Ed and Workforce Development, and with offices in Copenhagen, Leipzig, Massachusetts and Riyadh, Area9 clearly has its sights set on international growth.
We’ll wrap up with the words of Classting’s Dave Cho, who had this to say about the need for adaptive learning in the Korea Times:
"In the past, education was used to nurture laborers in the First Industrial Revolution. But industrial revolutions have continued, creating diverse jobs and requiring workers with many different qualities. But schools still teach the same things to everyone."
With so many well-funded global entrants on the case, the chances of a real breakthrough— the “Sputnik“ moment?— in AI-based Learning seem more likely than ever.