Thursday June 30, 2022

Not just another brick in the wall

December 01, 2021

The writer (she/her) is the technical adviser to the MoFEPT. Views are her own.

Every service in the world is turning towards personalisation – be it medical treatments, diet and fitness plans, shopping recommendations, streaming audio and video playlists, etc. The last domain where a one-size-fits-all factory approach is still tolerated is education.

In 1950, a few years after the end of World War II and only four years before he took his own life, Alan Turing developed the Turing test, an experiment for a machine’s ability to pass as sufficiently intelligent to make it indistinguishable from a human being. The test itself is simple enough. A human being, the machine under test and a human evaluator are put in three separate rooms. The evaluator conducts a natural language conversation with the human and machine in the other two rooms by means of written messages. Based on the conversation, if the evaluator cannot tell the human apart from the machine, the machine qualifies as intelligent. To date, no machine has passed the Turing test.

Last week Jensen Huang, co-founder and CEO of Nvidia, the company making a lot of the hardware and software enabling today’s Artificial Intelligence (AI) advancements, presented his company’s latest products and the applications they enable at its AI-themed GPU Technology Conference (GTC) November 2021.

There was an impressive hyper-realistic avatar (or virtual robot) indistinguishable from a real person on screen that could fool anyone, powered by Nvidia’s conversational AI named Maxine. It can deliver dialogue in any one of a number of languages while maintaining proper eye contact. Another one was Project Tokkio, a talking kiosk at a fast-food store, also powered by Maxine-AI.

In the medium term these and similar applications will let businesses replace millions of workers in customer service roles with AIs and avatars that will be available 24x7. Ultimately, however, restaurants, banks, retail stores are not in the business of developing AIs – they are in the business of making great food, issuing loans, and making investments and selling goods and services. Like a lot of other parts of their businesses, this tech intensive customer service will likely ultimately be contracted out. In this not-too-distant future, AIs will be subscription services for commercial and personal use. Want to enable the self-driving car feature on your car? Subscribe to an AI-chauffeur service for $1,000 a year.

It is always exciting to see upcoming technologies in action, but the demo that made me sit up and take notice was one of a conversational AI with a cartoon avatar of Huang himself placed in a hyper-realistically rendered virtual environment. Three people peppered it with complex questions and follow-up questions about climate change, astronomy and biology in various English dialects and the AI answered them all in real-time. The first machine to pass the Turing test could come any day now.

Think of the implications this has for education: The size of Pakistan’s under-19 population currently stands at 100 million. In time, the technology will be available to have effectively 100 million (AI) teachers for 100 million students. Like a personal tutor, each tailored to the personal learning level, able to pay attention and address the misunderstandings and misconceptions, able to challenge, to speed up or slow down the pace according to the needs of each individual child. An AI-teacher will not be constrained to teach all children a single curriculum or the minimum learning standards. The sky will be the limit, with the child in control of their learning pathway.

AI has already found significant use in delivering personalised learning. Duolingo, the most widely used language learning app, is a prime example that uses AI to track every error a learner makes and immediately adapts lessons and questions to focus on addressing weaknesses in understanding – an approach already proven effective in educational settings, albeit with real teachers.

We can already see instances where AI is supporting human teachers in classrooms. While it may be debatable whether AI-teachers can or should replace great, motivated teachers, they certainly should replace poorly performing, unmotivated ones. An AI-teacher has infinite patience and is guaranteed to not publicly embarrass students by unloading on them in front of their peers and extinguishing their self-confidence for getting an answer wrong or other sleight, something that is all too common. It would never tire and be available around the clock and could be trusted to not lie, be dishonest or be biased in favour or against any student for personal reasons.

All these advances in educational technology are heading in the same direction – personalised teaching and learning, away from the factory-school model where teachers teach to what the system considers the average student. Teaching to the middle leaves some students insufficiently challenged while leaving behind others who cannot keep up.

Meanwhile, here in Pakistan we remain committed to the factory-school model and are still far from deploying personalised learning strategies, not to speak of AI-teachers, even though we may have the greatest need for it.

You may think I am overemphasising the need for personalised learning but consider this: I recently had the opportunity to visit a community-based school. The school serves students who began their education late in life or were unable to join mainstream schools or dropped out of school. These are usually single teacher schools focused on providing students with basic literacy, roughly equal to a primary school education – the ability to read, write and perform basic arithmetic. The single classroom had children aged 5 to 16.

When I asked a 16-year-old girl (categorised as grade level 3) to read a line from a grade-2 textbook, she was unable to, while her teacher made awkward excuses. Just take a moment to absorb this. Think about that 16-year-old girl cooped up in a dingy room with at least 30 other children ranging from ages 5-16 and being unable to read at the level of a second grader. What must that feel like? I do not know why I asked her what she wanted to be when she grew up and that was the only time she smiled to respond ‘doctor’. I spoke with her mother (a sanitation worker) later and she seemed to be happy with her daughter’s performance. She was blissfully oblivious.

In the same classroom, I asked a 13-year-old boy (at grade 3 level) to add two 2-digit numbers that required a carry operation, which he was unable to do. For another student I pointed at a line in a nursery rhyme in her textbook, “Cobbler, cobbler, mend my shoe. Get it done by half past two,” and asked her to read it, which she did. When I asked her the meaning of the word ‘cobbler’ she answered ‘joota’ (shoe). She had memorised the line without understanding a word of it, but that was enough for her beaming parents who were happy to see that their child ‘speaks English’. Government statistics count all these as previously out-of-school children now put back in non-formal schools – an achievement. Yet, what future can any of these children realistically hope for in the modern world?

Our entire school system deserves more than constant fire-fighting – it needs a reboot. For the most part the entire school system – textbooks, teacher training, schools – is hopelessly outdated. On average, children that go to school for nine years stand where they should after four and a half years of schooling. The degree of learning disparity within (let alone across) schools and districts is so wide, it makes it impossible to tackle any problem only ‘one way’.

This is the era of bespoke-everything, and it is about time education was tailored to the varying needs of children in different school communities. That includes personalised learning pathways, possibly enabled by technology. The goal ought to be to let children reach their maximum potential, not just a mediocre but equal level. Achievement of a mediocre standard may have been a fine idea for the workplace two centuries ago (Read: Todd Rose’s ‘The End of Average’), but it is wholly insufficient for the twenty-first century future of problem solvers building systems and making great art with AI, quantum computing, biotech and other tech we cannot even imagine yet.

PS: Just as the world was slowly transitioning from pandemic Covid to endemic Covid, today we were hit by news of the spread of the B.1.1.529 Covid variant, labeled the Omicron variant, in South Africa and a number of other African countries. The WHO has designated it a ‘variant of concern’ while it is being studied for risk of reinfection and ability to escape prior immunity from vaccines. Until these questions are answered, there is a non-zero probability that parts of the world may go back to lockdowns, and for communities fortunate enough to have connectivity, students and teachers suffering from Zoom-fatigue from online schools.