How far has artificial intelligence come, and how soon will it reach online education?
In May 1997, Deep Blue, an artificial intelligence developed by IBM, defeated world chess champion Garry Kasparov, marking a significant symbolic advance in computing technology. Deep Blue accomplished this feat using a brute force algorithm, which allowed it to calculate thousands of possible moves in a short period of time and choose the ones most likely to lead to victory.
More recently, AI was put to the task of cracking the Chinese board game Go. Unlike Chess, which has a number of clear parameters for play, Go is an open-ended game, with only a few rules and a broad scope of play. Brute force algorithms proved ineffective, so instead, the programmers used neural networks so that the AI, AlphaGo, could teach itself strategy by running through tens of millions of games against itself. AlphaGo defeated the 18-time world champion, Lee Sedol, in March 2016.
While defeating grand masters at some of humanity’s oldest and most challenging games has raised the prestige of AI, it’s also been put to more practical use. In the field of medicine, artificial intelligence can now diagnose some diseases earlier and more accurately than doctors. Meanwhile, an pair of chat bots developed by Facebook to negotiate with each other quickly created a baffling code language to communicate.
As a new technology, it’s clear the full capabilities of AI have yet to emerge. It’s also clear that, as they improve and become more accessible, it will have many applications for online education. Here are just a few.
1. AI will help educators create new, immersive environments for scenario-based learning.
Right now, one of the biggest challenges to using branching scenarios in e-learning is that the number and complexity of options increases exponentially the further the scenario progresses. While these scenarios are immensely beneficial to learners, they can be difficult for educators to write.
However, AI could aid educators by generating dynamic responses within a scenario. Instead of following a pre-planned script, it could take more factors into account, creating a more realistic environment. For instance, the conditions in the scenario could change if a learner takes too long to respond, or if a certain scenario path loops back to a previous choice, it could remember those answers and cause them to change future options.
2. Learners will receive personalized tutoring sessions.
Online learning has given instructors a big advantage, in that it’s allowed them to create lessons that can be shared tens of thousands of times in massive online courses. This frees up time that the instructor can devote to creating more course material, marketing their program, or engaging with learners.
However, while the ability to automatically deliver content to a large number of learners at once has obvious benefits for running an online business, it can leave learners behind. There are limits to the amount of personal attention an instructor can devote to any student, and the larger that course becomes, the faster that limit is met.
Artificial Intelligence won’t be able to replace the roll of a tutor any time soon, but it can lighten the load. After all, if an AI can spot early warning signs of a disease, maybe it can also spot a learner about to fall behind in their lessons, and identify the most effective way to support them. Or, since we’re already designing advanced chat bots to handle customer support calls why not design them to help with language tutoring?
3. Automations will become more intelligent.
We’re always looking for better ways to automate common tasks. Repetitive duties, such as sending email reminders or grading multiple choice tests are already handled pretty thoroughly through pre-programmed workflows and set test answers. But some automations are too complex and contain too many variables and conditions for most of us to program reliably, and we haven’t yet developed an AI capable of grading a term paper.
But, strange as it may seem, the day may soon come when AIs can review short or long answers on a quiz and compare it to crowdsourced information gleaned from the Internet to test for accuracy. This would help teachers create more robust quizzes, without becoming overwhelmed by grading work.
4. More targeted and less intrusive marketing.
It may seem paradoxical to suggest that a more intelligent AI could deliver better targeted advertising and marketing materials without the need of much personal data from the customer, but it’s true. Many businesses, focused on big data, gather more information than they know what to do with. Because they can’t make heads or tails of this giant pile of unsorted data, they’re more likely to revert to the mean and send messaging that appeals to the lowest common denominator among their audience.
A better AI could identify which data points were relevant, and then stop collecting unnecessary data. That may sound disconcerting at first, but it’s actually a move in the right direction. Maybe it doesn’t need access to my Facebook account (along with my entire friend list, and their friend list, and so on) to know that I would be interested in an online language course. And maybe if companies learned how to target their advertising correctly, YouTube would stop showing me advertisements for rheumatoid arthritis medication.
Artificial intelligence still has a long way to go—but it is on its way.
Current AI technologies, while impressive, are still mostly limited by application. We can teach an AI to accomplish specific tasks better than humans, but we haven’t taught a single AI to do all the tasks humans can do better than we can do them. Deep Blue could defeat Kasparov at chess, but it couldn’t speak with him in Russian. AlphaGo might inspire some interesting philosophy papers on the nature of intelligence, but it can’t itself write them.
However, the improvements in neural net AI that helped crack Go are pushing boundaries in all kinds of industries—education among them. As much as these scenarios may seem to belong to the realm of science fiction, it’s likely many of them will be commonplace within the next 5–10 years. Will you be ready?