Insights From The Blog

AI is Changing How We Learn and Create

Love it or loathe it, AI is now firmly on the table. Already, some writing-based websites are not only warning that AI content will be banned and major stock photo repositories are stating that AI-generated content must be marked as so. Driven by a mainstream media frenzy, AI seems to have become the bad-boy of creativity.

But AI is something to be embraced, not feared. Despite the continual media coverage of how intelligent computers are, it is critical to recognise that AI does not contain much intelligence. The capacity for logic, comprehension, self-awareness, learning, emotional knowledge, planning, creativity, and problem-solving is referred to as intelligence. However, machines can only do a few of these tasks right now. The ability to think is implied by intelligence. However, AI does not think in the conventional sense. 

The fundamental power of AI rests in the way it processes massive amounts of data and then makes educated assumptions in terms of decision as a coherent – and seemingly intelligent – output. The more computers repeat the same task over time, the more they can alter their information processing and, as a result, the better they become at guessing. This is essentially how machines ‘learn’ information and disseminate it out.

One area where the power of AI is really starting to have an impact is that of education and the delivery of educational content. This isn’t, however, driven by the notion that ‘it’s inevitable that it will happen, so let’s just accept it’, but rather by an increasing amount of hard evidence.  AI learning and tutoring may well be in its infancy, but the results have shown that, even in its current state, people tutored by such a system perform better in tests and exams. 

Educational physiologist Benjamin Bloom – he of the triangular ‘Taxonomy of the Cognitive Domain’ – has also developed the notion of the 2-sigma educational standpoint. In this, Bloom asserted that a student who was schooled on a one-to-one basis by a suitably prepared tutor was likely to gain 2 standard deviations higher in results than those schooled in a multi-educational environment. Two standard deviations in this case would be the difference between getting a ‘C’ and getting an A, or a ‘B’ and an ‘A+’. That’s a pretty big difference.

At the time, Bloom focused on individuals and groups who were tutored by humans but in different settings, and found the disparity. Now, with AI tutors becoming a reality, we are likely to see not only an increased incidence of this but the possibility of increased sigma range – a ‘D’ to an ‘A’ is perfectly possible.

Generative AI isn’t that new, but we are now coming to terms with what it can do and where it is leading us. Researchers found that AI systems exposed to huge amounts of data gave much better results than less well-informed systems. By 2014, software systems were starting to make much better sense of language and started to obtain a natural feel. By 2017, these language models were being consolidated and research was focusing on the actual performance of the AI systems and understanding how these could be enhanced by additional information input. 2022 saw the launch of ChatGPT and allowed users access to AI with a huge conversational ability. This model has given way to a range of tutorial bots that have instructional abilities. Simply put, AI has been trained over the last decade or so to become an expert at natural language conversations.  But more than that, it has sufficient ‘intelligence to operate as a tutoring tool. Suddenly, everyone can have a one-on-one tutor, and following the work done by Bloom, this means that everyone’s grades should go up.

Obviously, generative AI isn’t the same as a human teacher, and it remains to be seen whether an AI personal instructor can mimic the learning incentive and efficacy that comes from human connection with added empathy. However, the potential for collaboration between AI and human tutors to boost both tutor productivity and learning impacts for students is likely more exciting. In any event, it’s evident that we’re on the verge of AI-enhanced teaching and learning, and these developments are happening much quicker than human behaviour can adjust. The question becomes one of whether humans can accept the tutelage from an AI tutor, or will there be some kind of resentment or air of disbelief with such lessons.

In the first Star Trek movie, we are treated to a scene where Mr Spock as a young boy is being tutored along with many other Vulcan children by AI tutors. Compared to where we are now, that doesn’t actually seem to be too much of a stretch of imagination. But to reach that, nicely antiseptic environment, we need to embrace Generative AI, and that acceptance becomes a problem.

Many commentators think that AI needs to be introduced in more well-accepted ways, such as via gaming. By gamifying AI we learn to not only accept it, but ultimately embrace it. Software such as Unity has a pretty steep learning curve, but tests have shown that by introducing generative AI into the learning platform, new users pick up the fundamentals quicker. By giving the training of things like software packages over to AI, more people will learn to embrace the concept more quickly. Add AI to online training packages and before you know it, it becomes an accepted as the main way of taking up information. 

Unity, as a technology company, is committed to the use of generative AI, and we think that’s good news. AI has already shown itself to be a powerful tool, and if it can be weaved into the world of learning, then it’s believed that more people will seek out knowledge delivered in this way.  When that happens, we predict an avalanche of learning, and that has to be beneficial for everyone.