In the past, the non-linear acceleration of computing power and artificial intelligence (AI) offered an unprecedented opportunity to society to achieve learning outcomes. In concert with standards, governance, and human intelligence, AI is navigating its way towards an equilibrium state in the education domain.
In 1956, a small group of scientists gathered at artificial intelligence (AI) research workshop held at Dartmouth. That conference, from over 60 years ago, is widely considered the birth of AI research. Over the ensuing years, AI has moved from the fringe of technical research to become the centerpiece of innovation. The advance of AI critically depends on computing power, algorithms, and data, all of which Learn with AI have experienced tremendous growth. Take, for example, computing power: the performance of the world’s fastest computer accelerated from 124 GFLOP/S in 1993 to 148M GFLOP/S in 2019, i.e., over a million times over the past quarter-century. And in specific fields such as image and speech recognition, AI has caught up with human ability by demonstrating an admirable ~5% error rate. Today, the increased maturity of AI has enabled rapid adoption across industries and is transforming education.
In primitive society, knowledge was passed on orally by elders to their children; however, around 2076 BC to 1600 BC, the Xia Dynasty created the first formal schools in China. Since then, education has been defined as an exclusive relationship between teachers and students, and its success—and the rate of knowledge transfer—hinged heavily on the teacher’s competence and the student’s ability. The growth of AI is changing and augmenting the teacher-student relationship in a number of ways. First and foremost, the standard for curating learning telemetry data is being redefined to ease the AI training. In the past, teaching best practices that were largely based on individual knowledge are now gathered and standardized to improve outcomes for less experienced teachers. Further, AI is curating and injecting predictive and prescriptive insights, going beyond descriptive analytics, accelerating mean-time-to-detection (MTTD), and mean-time-to-correction (MTTC) of learning opportunities; and learning pathways are individualized to improve the student experience and speed up the delivery of outcomes. Finally, human-computer interactions are naturalized through haptics and dialogue to reduce friction in the learning process.
How will AI impact teachers’ jobs? A close analogy is a fear not long ago that bank ATMs would make tellers obsolete. As a matter of fact, the number of tellers increased from 300K in 1970 to 600K in 2010, and automation is credited with enhancing productivity, which enabled banks to open more branches. In much the same way, AI’s role in education will continue to grow by freeing up teachers from repetitive and clerical tasks so they can focus more on core education services such as teacher-pupil interactions and the creation of high-quality educational content. One can predict that the AI-teacher symbiosis will continue to evolve even further, and future generations will have to work to ensure that AI is applied fairly.
This leads us to the inherent bias in AI and its social justice ramifications. Much like the humans who create machinebased intelligence, AI is prone to bias: from the humans who train AI; in the data that is used to train the AI; and in the AI that trains another AI. These inherent biases raise important social justice issues in education, and it is our collective responsibility to recognize these limitations and collaborate to enact policies that are transparent, equitable, and sustainable.
For the foreseeable future, man and machine will co-exist, but we need to ask ourselves: while AI machines are being trained and learn about us, won’t we (as a society) be interested to know how they (machine) view our world? Education will be a central testing ground for AI innovation, testing the limits of these questions and others to help shape the future of AI and the human-AI symbiosis.