For the last six months, I’ve been saying to people that the rise of Generative AI is the biggest technology shift that has happened since the launch of the iPhone and, before that, the launch of the Internet. In ten years’ time, we’ll be sitting in a bar, sipping our martinis (shaken, not stirred, of course) and saying, “Do you remember when ChatGPT went mainstream and everything changed?”
The vast majority of these changes are going to be for the better, and I am super excited about the thousands of companies out there pioneering new products and features off the back of the recent breakthroughs in generative AI. But one of the big losers will be the traditional education publishers. For these companies, content is everything.
When I started my Ph.D. in Robotics and AI at Oxford back in 1992, I was excited about the changes that AI would bring, but it was not until this year, 31 years later, that I felt that AI had finally come of age and was about to truly transform the world.
What is Generative AI?
Generative AI, short for Generative Artificial Intelligence, refers to a branch of artificial intelligence that focuses on creating or generating new content, such as images, videos, text, music, and more. It involves using machine learning techniques, particularly generative models, to produce original and realistic data that resembles the patterns and characteristics of the training data it was exposed to.
I used ChatGPT to generate that paragraph above. It generated a lot more too, but for me, the best way to describe Generative AI is like this … imagine that you had a personal assistant with a photographic memory of all the world’s publicly shared content (text, images, audio, and video) and who was highly skilled at taking a high-level description of some content that you would like to create and would then create it for you in seconds. Importantly, your assistant won’t just copy and paste what it knows about; your assistant has the ability to understand language and context and will generate something new and original that fits your description and needs exactly.
ChatGPT 3 was trained on over 45 terabytes of text data. The entire works of Shakespeare take up only 5 megabytes, so ChatGPT was trained on the equivalent of 10 million times the amount of content in the entire works of Shakespeare. That’s quite a lot of information!
Generative AI for teachers
Imagine you were a teacher in a classroom; perhaps you would ask that assistant to:
- Write open-ended questions: “Give me 10 question prompts to assess students’ understanding of animal adaptations.”
- Generate descriptions of historical events and biographies: “Summarize the life of Anne Frank in 50 words or less.”
- Design a unit outline: “Develop a 5-day unit plan outline for teaching students about ancient civilizations including Aztec, Roman, Greek, Chinese, and Mayan.”
These are just a sample of some of the 50 “prompts” listed on this page, which in itself is just a fraction of the many thousands of prompts that teachers might enter into ChatGPT. ChatGPT will do an admirably good job of creating content for all of them. It’s far from perfect and will sometimes make mistakes, but overall, it’s remarkably good.
So here’s the problem for the publishers: why would a school, teacher, or student pay money for publisher content when they can generate it for free on-demand using generative AI? More importantly, the content that the teacher and student generate can be instantly customized to their specific needs and the needs of their student. Perhaps they can create the content with a prompt ending in “… for students that are struggling with multiplication” or “ … in language suitable for a student with an age 5 reading ability” and many more refinements.
Instant, free, customized content perfectly suited to your specific teaching needs and to the needs of your students. It sounds like something for the future, but the reality is that it is already here today.
What does this mean for publishers? Smart publishers are realizing that the only business models that make sense for the future are those where content is free and where they embrace Generative AI. Whilst publishers will still be able to make money from content in highly specialized areas where the learning content is unique and similar content is not available publicly on the Internet, there will be no money in mainstream content. Business models need to change for publishers to succeed. The question is how?
Classroom engagement and interactivity
The pain for publishers doesn’t end with how generative AI is allowing teachers to instantly create high-quality and customized content instantly. The way teachers are teaching in the classroom is changing too.
Classrooms are now far more interactive, and as a direct result of the covid pandemic, one-to-one student devices in the classroom (typically Chromebooks) are now the norm rather than the exception. Nearpod, Peardeck, Kahoot, Quizizz, and our own product Quizalize have become the teacher tools of choice for delivering content in the classroom and for homework. They enable teachers to deliver content in a far more engaging and interactive way. Gamification, too, has become mainstream, with Blooket and Gimkit demonstrating the cutting edge of classroom gaming, and we believe our own Minecraft-inspired game Blockerzz has set a new benchmark for multiplayer classroom quiz games.
All the above companies (including us) have focused on UGC content models for two good reasons. First, there is no need to invest in creating content yourself if your users create it within your product. This allows such companies to invest in deep technology instead of creating content. Second, with a UGC model, your product can be used in any market around the world automatically. As a result, these companies have global traction. However, The one challenge for the above companies, historically, is that it takes teachers time to create content themselves. Indeed it typically takes a teacher around 30 minutes to create an interactive presentation or quiz to deliver to their class. The most active users of those products to date, and the ones that pay, have been those who have been willing to put personal effort into creating such content. Whilst these products are now mainstream and a staple of the US classroom, the effort required by teachers to create content is still a barrier to much broader and frequent usage.
However, the arrival of Generative AI means that teachers can now create content for these platforms in 30 seconds rather than 30 minutes. It means that not only can they create exactly the content they want automatically in seconds, but they can also deliver it in the classroom in an engaging and interactive way that students love. Even better, by delivering interactive quiz content live in the classroom, teachers get real-time data that they can use to customize and personalize their teaching further.
For the companies above, it’s a complete game changer for adoption. Any teacher can now create exactly the content they want in seconds. Moreover, it moves using these products from being a pre-planned activity requiring 30 minutes of preparation to being an activity that can be done in the classroom on demand whenever needed. Teachers have been blown away by our ChatGPT-powered AI quiz creator, yet we are only just getting started. We have some innovative new features coming shortly in Quizalize that will take the formative assessment to a level of interactivity, features that are only possible as a result of Generative AI.
The big publishers have been slow to embrace this new generation of interactive classroom experiences. For those that have gone digital, they have largely focussed on staid and boring learning products that schools buy and use because of the high-quality branded content inside rather than for the teaching and learning experience they provide. However, now that any teacher can create high-quality, customized content on demand for free in seconds, it also means that they are now freed from the shackles of having to use a publisher’s digital product and instead are free to use the interactive and engaging products that both they and their students love.
Differentiation, personalized learning and virtual tutors
I formed my own company, Zzish, with the mission to “give teachers the superpowers to personalize their teaching to every student,” and Generative AI now means that we can take a giant step forward to delivering on this mission. As we have discussed, Generative AI allows teachers to create more custom content suited to the needs of their students, but it’s now so quick and easy that it also means they can now create different content for different groups of students that they are teaching customized to their ability and level of understanding.
The best teachers we work with are already making an effort to create and deliver different content for the different groups of students that they teach. This is known as “differentiation,” and different content is assigned to different groups of students in the classroom depending on their current level of understanding of a topic or learning objective. Differentiation is a little weaker than full personalization, but it is still hugely impactful. Our own product, Quizalize, is unique in its category at providing built-in features for differentiation, and teachers who use it often tell us how powerful they find it. However, it takes a significant amount of effort on the part of the teacher to provide the content to differentiate their teaching. Generative AI transforms this and now makes it possible for teachers to do this in a fraction of the time that it took before. Differentiated teaching is becoming far, far easier.
For traditional publishers, this presents an extra problem to compound the issue of Generative AI. Whilst publishers create very high-quality content for teaching, the huge effort and investment it takes for them to create this content means that it rarely includes resources for differentiation. Indeed this content is usually created with a one size fits all approach in mind.
However, helping teachers produce more customized content doesn’t end with differentiation; Generative AI offers up the opportunity for full content personalization. The Khan Academy and Duolingo have both launched “virtual tutors” based on ChatGPT 4, the latest version of ChatGPT. Indeed ChatGPT 4 takes Generative AI to an entirely new level and makes the prospect of every student having their own personal virtual tutor a genuine possibility. Our own experiments with this latest technology have demonstrated just how powerful this technology is. Teachers that have been testing our prototype have been blown away by what it can do.
It’s not quite there yet, though. Chat GPT and other similar products like Bard still makes mistakes, and, unfortunately, they make those mistakes “authoritatively”. I’ve had ChatGPT give me a very thorough and detailed explanation of why 2/3 and 3/5 are equivalent fractions, for example. However, we are still in the earliest stages of this new technology. We are perhaps at the iPhone 4 stage of maturity, and over the coming years, the quality of these products will improve dramatically. I have no doubt that we will reach a level of quality very soon to deliver virtual tutors who give authoritative, detailed, and customized explanations that even an experienced teacher might struggle to deliver.
What is absolutely certain is that Generative AI, and large language models like ChatGPT in particular, are transforming the education landscape. There will be new winners and new losers. Publishers need to embrace it, or they will be left behind. To win, my recommendation is that they take the following three steps:
- Develop new business models centered on the core principle of mainstream content being free.
- Leverage generative AI in-house to help authors create more varied content for differentiation and personalization.
- Invest in and partner with companies that deliver classroom content in new interactive, and engaging ways.
I’m sure the publishers will argue against the need to make their content free. They will argue that the quality of their content is far superior to that produced by tools such as ChatGPT and that ChatGPT can’t put together pedagogically sound material. This is indeed true today, but as this technology matures, this will change. Moreover, it partially misses the point. Teachers actually want to create their own content as they want to customize it to their specific needs and apply their own pedagogy; it just takes them too long to create that content themselves from scratch at present. ChatGPT makes it far more efficient for them to do this. Moreover, it actually doesn’t matter if ChatGPT gets it wrong sometimes as long as the teacher is able to review and correct it easily. Indeed, our own tool Quizalize learns from these teacher corrections to produce an ever-improving layer on top of ChatGPT itself.
For publishers, the big discussion is no longer about digital vs. print; the world has moved on. The battlefield has moved to how content is delivered digitally, and the battle will be won by those companies who successfully deliver content personalized to each student’s individual needs in an engaging and interactive way that has a real impact on helping students progress.
Generative AI is a real game changer for education.