Intelligence seems to be making a comeback. No, we’re not talking about human intelligence — that one is still out of fashion if we’re to judge from the news. Rather, it’s Artificial Intelligence that appears to be back from its long hiatus, and this time it looks as if it’s here for good. It’s already an investment focus in Silicon Valley, it draws large sums of R&D money, it powers many exciting new startup ventures, and it promises to revolutionize, well, just about everything.
In this post, we are going to examine Artificial Intelligence’s past history and latest developments, discuss its future trends, and attempt to determine the ways in which it will affect modern corporate training.
Back with a vengeance
In case you’re not familiar with the history of AI, you might be wondering what we meant when we said that it’s “making a comeback”.
After all, isn’t AI a new, or at least a relatively recent, development, pioneered by such technological advances such as IBM’s Watson, Apple’s Siri, Microsoft’s Cortana, Google’s TensorFlow, and the like?
The answer is no, it isn’t. It’s actually older than The Beatles, with relevant research starting sometime in the mid-50s, and reaching its highest point in the late ‘60s and early ‘70s. It’s rise was put to an end, though, during the so-called AI-winter, when (after the grandiose promises of early AI researchers failed to materialize) exploratory research funding was abruptly cut off.
Fast forward forty years and computing advances later, alternative research approaches such as deep learning, and new business models are making AI research relevant once more.
So relevant, in fact, that not only are billions in research money poured into AI research, but its potential has even have some (otherwise very bright) people scared for our future.
So what does that mean for corporate training? We’ll find out, but first…
A (side)note on terminology
For some people Artificial Intelligence translates to “thinking machines”: digital or software based “brains” that are capable of human-like reasoning.
That (usually called “strong AI”) is not the only way to think about the problem, though. A program that can play chess, or recognise faces in photographs is also worthy to be called an “artificial intelligence” even if that’s all it can do.
When the industry uses the term “Artificial Intelligence” nowadays it means mostly of the latter kind (task specific intelligence) ― the former (general intelligence, i.e. “thinking machine”) is still many decades away.
eLearning in general, and corporate training in particular, have traditionally been quite plain affairs.
What has been revolutionary about online training is not its smarts, but rather the cost savings, learner convenience, widespread reach, and operational simplicity that it brings over classroom-based training. At the end of the day, though, online learning is still about uploading some content and having learners study it.
Will it be able to replace eLearning instructors? Maybe — in the long run. But it already has the potential to greatly assist them in their daily tasks and enhance training program effectiveness.
Let’s see how…
While the idea of “learning styles” has been discredited in recent pedagogical research, what’s indisputable is that different learners have different strengths and weaknesses in their understanding of particular material.
The best traditional teachers pay special attention to the needs of each individual student in their class — they encourage the brighter ones to study further, attempt to explain a concept in simpler terms to students who find it challenging, hand out personalised assignments, and so on.
This becomes increasingly difficult as traditional classes grow in size, and nigh impossible in online training settings, where an instructor might be responsible for hundreds or even thousands of learners.
This is where artificial intelligence algorithms can be used to “cluster” learners based on the abilities or weaknesses they display (as determined by an aptitude test or, even better, by an incremental assessment of their training performance), and assign appropriate material, exercises and tests to them.
Kind of like eFront’s existing Skills-Gap testing feature, but on steroids. Beyond identifying missing skills and recommending courses that offer them, AI systems would automatically adapt the content of a course to the needs of the learner.
As research into “emotional intelligence” of software agents progresses, such an AI might also be able to employ strategies that connect with the learner in an emotional level, e.g. by sensing their discouragement, or by playing into their personality traits (e.g. their competitiveness) to increase their engagement.
Using AI to analyze a learner’s strengths and weaknesses (as discussed in the previous section) has even more immediate applications than personalized courses.
It can be used, for example, to create a profile for each learner that goes beyond their skill-set (or lack thereof), and provide fuller insight into their learning abilities, aptitude for particular skills or job positions, and determination.
Plus, by plugging in even more sources, like job performance data, it will be feasible to get a comprehensive profile of the employee that extends beyond their training and into their overall productivity.
Such an AI tool, hooked with the company’s ERP and Learning and Talent Development Platform, could be part of a larger system that tracks employees all the way from pre-employment assessment and on-boarding, to induction training and everyday performance.
Creating a good training curriculum is hard.
Well, not the content creation part: eFront makes that one easy. The “coming up with good content” part, though, is. And it’s equally hard to evaluate your training curriculum based on its actual success in training your learners (or, even better, the kind of ROI it brought your company).
Some platform features, such as Reporting and Surveys, make this job easier, but wouldn’t it be better if the software could assist you even more? In the near future, artificial intelligence systems should be able to do just that.
A lot of necessary ingredients are already available: we have machine learning systems capable of analyzing large volumes of data, and finding relevant correlations and inferences. We also have, courtesy of xAPI, access to streams of training activity data that cover all kinds of learner actions across both the desktop, browser and mobile space.
Putting the two together is a no-brainer, and with the right algorithms it will result in expert systems that are able to examine training activity data, find weak spots in any course or curriculum, suggest improvements, and, in some cases, make automatic adjustments that increase a training program’s effectiveness.
Automatic Content creation
Will a machine ever be able to create new training content?
It might, but don’t fire your content writer just yet.
See, this is where we need to jump from task-centred AI (which has great promise and has been effectively put in production in all kinds of contexts) to strong(er) AI (which is still quite elusive).
Creating content requires an actual (and full) understanding of the subject matter, and that cannot be replicated by the kind of machine learning or deep learning algorithms that dominate today’s AI research.
That said, given a large corpus / pool of curated and tagged training content, AI tools can help create large variations of it, weighted and focused on particular training needs.
Kind of like eFront creates new Tests and Quizzes from an existing pool of questions, but with extra smarts to be able cater for specific learning needs and move beyond randomly combining pre-existing questions.
Even today, for specific domains, such as mathematics and language learning, simple AI rules can produce an almost infinite variety of test questions (e.g. automatic creation of n-th degree equations).
There’s also automatic content translation, which, although far from perfect (try Google Translate with anything even slightly challenging), has the potential to save multinational businesses millions of dollars and great increase their time-to-market. Though, judging from the quality of some product manuals I’ve seen, this technology is probably already in widespread use.
An intelligent Summary
In this post we had a look at Artificial Intelligence ― what it is, how it came about, what it promises for the future, and in which ways can it help improve eLearning in the here and now.
Indeed, “improve” is perhaps the key word here, as we don’t expect to see revolutionary changes in the short- and mid-term, but rather an increased application of AI and machine learning concepts in eLearning that will help automate menial tasks and increase the overall effectiveness of employee training.