Tech Tutor
Exponential technologies are changing how and where we learn
Exponential technologies are changing how and where we learn
The Virtual Reality (VR) module nudges you to use a controller and point towards the image of a brain, suspended in the air. The brain, seen through a VR headset such as Google Cardboard, can be rotated to get a 360- degree view. Dramatically, it splits into separate lobes. Click on the lobes and you will be told what each one does. The Occipital lobe, for instance, is associated with visual processing. The split brain can then be rejoined; you win points for putting the right lobes in the right places. The simulated experience is far more absorbing than reading from a text book. Veative Labs, the creator of the module, says such simulations close the gap between knowledge and understanding, converting passive learners into active ones. The company’s science and mathematics modules are being used to teach complex concepts in classrooms across 25 countries. “It is easy to see that we augment the learner’s ability to learn. What is less obvious and just as true is that we augment the teacher’s ability to instruct by expanding their repertoire and resources,” the company states on its website.
Mostly used by elite private schools so far, the adoption of
VR and augmented reality ( VR) for better learning outcomes has now entered conversations around the proposed new education policy. India’s Draft National Education Policy 2019 states that “technology use and integration will be pursued as an important strategy for improving the overall quality of education”. A variety of educational software would be developed and made available for students and teachers at all levels. Among them would be “serious games, simulations and applications using AR and VR”.
This underlines the way technology is challenging the traditional notions of teaching and disrupting the classroom model. Teachers teach from homes and students consume content and lectures from anywhere using phones. Student engagement cannot be tracked scientifically in an offline classroom. This is changing. Artificial Intelligence (AI) and its branches such as Natural Language Processing ( NLP uses software to process and make sense of human language) can gauge a student’s comprehension abilities as well as analyse sentiments during live online classes. Face recognition technology captures both student engagement and teacher effectiveness. Commercially available AI tools help students identify the colleges and courses they are suited for.
“The biggest thing happening today is AI and specifically, within AI, machine learning and deep learning,” says Sebastian Thrun, the Founder, President, and Executive Chairman of Udacity, a Silicon Valley- based unicorn that offers massive open online courses.
“AI can be used to make a world- class teacher. World class teachers don’t just lecture — they listen, understand where the student is, and are able to adapt to the student. We will see an AI that will make education adapt to the students,” he adds. Udacity began as an experiment in 2011 when Sebastian Thrun and Peter Norvig, both instructors at Stanford, offered their “Introduction to Artificial Intelligence” course online for free.
“The core is how students can learn better. Udacity uses AI to figure out which part of their education is broken. How can students be more efficient in learning? We also use AI for pricing decisions, the right price point for a market and the right products. It is similar to the way Amazon uses AI,” says Thrun, who is also the founder of Google X ( projects included the Self- Driving Car, Google Glass).
Thousands of Indian entrepreneurs have also dabbled into education technology, Edtech in short. According to DataLabs, the research wing of Inc42 Media, 4,450 edtech companies started in India between 2014 and 2019; about 1,150 of them shut down. India, today, has 194 funded edtech startups. The overall funding pooled in by the sector is nearly $2 billion. Many of these start-ups have innovated around live tutoring. A tough ask because of bandwidth constraints in many parts of India.
Optimising Live
Much like Thrun, Vamsi Krishna was a teacher too. He taught at Lakshya, an offline coaching company he co-founded with friends in Punjab, in 2006. By 2012, when he exited the business, Lakshya had become a well-known test preparations brand in North India. He sold the company because it was getting difficult to scale up
“UNIVERSITIES GIVE A LOT MORE WEIGHT TO PEOPLE WHO WORK IN START- UPS. WE KEEP CHANGING THE WEIGHTS SO THAT THE ALGORITHM IS UPDATED”
Akshay Chaturvedi
Founder and CEO, Leverage Edu
offline centres. “We figured out that expanding an offline centre is not easy. Replicating the soul of one centre, maintaining the same quality of teachers, is tough. If impact at scale is the goal, it has to be tech- enabled,” he says.
Krishna next co-founded live tutoring company Vedantu to teach students between Classes VI and X all subjects. In its formative years, around 2016, the company facilitated one- on- one classes (where one teacher taught one student). On an average, one teacher now teaches 200 students. Vedantu has a freemium model; in its free classes, the number of students per teacher can go as high as 2000. That’s the sort of scale technology enables.
“We have developed a live streaming technology. The closest analogy is a Zoom call or a Skype call. The difference is that ours is custom-made for teaching and learning,” says Krishna, sitting in his office in Bengaluru’s HSR Layout. The wall of his conference room is inspired by Star Wars. Key to the company’s success with live teaching has been the optimisation of video streams. In low network areas, the streaming isn’t interrupted. Here’s what happens: there are many elements to what is streamed. There is teachers’ voice, their video, and scribbles on a writing device. At the student’s end, the scribbles appear on a whiteboard. While the voice and the video are streamed, the whiteboard isn’t. Vedantu has the controls to the scribble – only what is written is transferred and then transposed on the whiteboard. “The size decreases massively. We are saying that the board is static and what is changing there is just the scribble,” says Krishna.
Another company optimising live streams is Unacademy, an educational technology company into test preparation. Unacademy started in 2015 with an app that made content creation easy for teachers. “People liked to share their knowledge but the barrier to entry was high. Teachers needed a mike to record audio, Power Point for presentations, and editing software. Then they had to upload it on YouTube and worry about distribution,” says Hemesh Singh, Co-founder & CTO of Unacademy.
The company’s app has simplified the process. Teachers can now record video using the phone’s in- built camera, upload Power Point presentations, highlight important parts in the text and do voice- overs. The videos and other visual content are uploaded on Amazon storage servers before being delivered to students. “Many competitive exam toppers started teaching online because it was so easy to make it,” says Singh.
Unacademy currently has more than a million videos created by over 10,000 educators on its platform. Recently, the start-up announced a funding round of $110 million. It counts Facebook, General Atlantic, Sequoia India, Nexus Venture Partners, Steadview Capital and Blume Ventures
among its investors.
Matchmaking
The LinkedIn description of Leverage Edu is crisp but doesn’t tell the whole story: “The primary platform is a marketplace that connects students with experts ' just like them' using machine learning, to help them with career guidance, college applications, get job ready, and more”.
At a noisy café in Delhi’s Lodhi Road, Founder and CEO Akshay Chaturvedi is in a mood for more detail. About 7,50,000 people visit his website every month. An AI tool helps people understand the right foreign universities they should apply to and the programmes that are best suited for them. How does Leverage manage this?
The company’s matchmaking algorithms are fed with tonnes of data. “We trained the machine with a decade’s data — of people who applied to different kinds of schools and the programmes they chose,” says Chaturvedi. The algorithm also factors in the background of the person looking for a university. Inputs include what and where she/ he studied previously, Graduate Management Admission Test scores, work experience, if any, and the employer. Different inputs have different weights. “A lot of feedback is taken from the universities as well. Universities today give a lot more weight to people who work in startups, for instance, versus whose employed at big firms. We keep changing the weights so that the algorithm is updated,” says Chaturvedi. The algorithm suggests between three and five universities across the world. “We are 80- 85 per cent accurate. Since the machine is not always 100 per cent right, there is an offline counselor for people to talk to,” says the founder.
Leverage also uses AI to predict career choices. “We do this as part of some workshops. The algorithm is more clinical psychology- driven. It factors in how you respond to different situations,” he says. The machine evaluates the answers based on mini case studies that the candidates respond to. For instance, it can tell if one has a marketing bent of mind or is more finance- oriented.
Meanwhile, machines can also tell if those seeking career counseling from mentors online had a useful conversation; if they were attentive; if the sentiments were expressed loudly. When it comes to live classes involving younger students, machines can sense their comprehension abilities. For example, are they thinking through what they read or hear?
Analysing Learning
Any future of work expert will tell you that critical thinking is probably the most important skill. Machines, going ahead, will do most of the repetitive work, leaving very high order cognitive stuff for humans.
AugLi, a Gurgaon- based company, has built a platform that tests critical thinking among children between 10 and 16 years. The platform is modelled around current affairs. It gauges a student’s reading age versus his biological age before pooling in content appropriate for her/ his level, everything from sports to physics. It could be an article about Mangte Chungneijang Mary Kom, the Indian boxer. After reading the full article, the student
is nudged to summarise it through either audio or text. “The ability to identify what is important in an article and assimilate is critical,” says Anjali Tiwary, the CEO and Founder of AugLi. “We tell you if you have comprehended the article well enough. If you recorded the summary, did you speak at an appropriate pace or were you too fast?” The “We” in Tiwari’s statement are not humans. Algorithms figure out how well one has understood what was read. The feedback is possible because of NLP, a branch of AI.
This is how it works: In case of an audio summary, the audio is first converted into a text transcript. The two documents – the original article and the smaller summary – are then compared by the algorithm. “First, nouns are taken into account and matched. Next, the phrases in the two documents are mapped. If the student has used terminology not in the original document, the machine opts for Ontology (organising information and interpreting interrelationships between ideas),” says Kamal Kashyap, Co- Founder, AugLi. “The system has access to thousands of other documents. It, therefore, understands the relationship between different concepts.
“TEACHERS NEEDED A MIKE TO RECORD AUDIO, POWER POINT FOR PRESENTATIONS, AND EDITING SOFTWARE. THEN THEY HAD TO UPLOAD IT ON YOUTUBE AND WORRY ABOUT DISTRIBUTION”
Hemesh Singh Co-founder & CTO, Unacademy
When the user produces something not in the article, we are able to map it out,” he adds.
Vedantu also used data science and NLP to understand students and how they are learning better. Face recognition technology is used to gauge student engagement and teacher effectiveness. Is the student looking at the screen or looking away? Are they smiling or frowning? Are they blinking too much? What’s the pupil dilation? What is the time delta between the teacher asking and the person answering? In a live class, even the tone of delivery is analysed. “We capture over 70 parameters from every class. We have unprecedented data to understand where the child is in terms of his understanding,” says Vamsi Krishna, Co-founder of Vedantu. “Using this data we can create customised content. Live learning analysis is where our strongest patents are,” he adds.
The Draft National Education Policy 2019 talks of similar goals for every school. Software must be developed to help teachers create adaptive assessments that provide appropriate feedback to learners. “Such assessments will minimise the importance of rote memory, and will instead focus on 21st century skills, including critical and creative thinking, communication, and collaboration,” the ambitious draft policy states. Impact at scale is the goal.
“WE TELL YOU IF YOU HAVE COMPREHENDED THE ARTICLE WELL ENOUGH. I F YOU RECORDED THE SUMMARY, DID YOU SPEAK AT AN APPROPRIATE PACE OR WERE YOU TOO FAST?”
Anjali Tiwary, Founder & CEO, AugLi; Kamal Kashyap, Co-founder, is in the background