top of page

AI State of the Union: A Look Back and A Look Ahead for Education

ChatGPT recently turned 1 year old and in that last year, and since then we have seen a Cambrian explosion of new large language models or LLMs as well as 1000's of applications that are leveraging the compute and inference power of this kind of AI. Even when compared with the rapid proliferation of technology in the last decade, the last year has felt like things are moving exponentially faster because of AI. Once a niche technology, AI is poised to have to be a significant disruptor when it comes to everything from entertainment, to how we work, to teaching and learning and just about everything in between. Understanding where this technology has come from and where it is going can help you stay ahead of the curve, and so in this post, we will look at the story arc for AI from its current everyday uses to what we might see in 2024. Buckle up, and let's go!

Current AI Tools in Everyday Use

While platforms that allow us to interact with AI models like ChatGPT, Google Bard, or Claude are rather new, the use of AI in our lives isn't. The reality is that many people are using some form of AI almost every day. Some of the more common examples include: Our daily lives are increasingly intertwined with AI, often in ways we might not immediately recognize. Let's delve deeper into some of these instances:

  1. Recommendation Algorithms in Entertainment: Streaming services like Netflix and Hulu utilize sophisticated AI algorithms to analyze our viewing habits. By examining what we watch, when we watch it, and even what we stop watching, these platforms can offer highly personalized content recommendations, thereby enhancing our viewing experience. As an example when you check out shows or movies on Netflix, you may see a matching percentage, where the AI has determined what it thinks will be a match with your tastes. Recommendation engines can also be found in music streaming services as well.

  2. Language Translation Tools: AI-powered translation tools like Google Translate, have broken down language barriers that once hindered global communication. These tools use natural language processing (NLP) to understand and translate text or speech between languages in real-time, making international travel, business, and cross-cultural communication more accessible.

  3. Navigation Systems: AI's integration into navigation systems like Waze, Google Maps, and Apple Maps has revolutionized travel and commuting. These systems analyze vast amounts of data, including traffic patterns, road closures, and even weather conditions, to offer real-time route optimization.

  4. Voice-Activated Assistants: Devices like Amazon's Alexa, Apple's Siri, and Google Assistant have brought AI into our homes, responding to voice commands to play music, set reminders, or even control smart home devices. These assistants continuously learn from our requests and preferences, becoming more efficient and personalized over time.

The examples of the technologies above are all a good bit older than the current AI technologies that have been making headlines in the last year. The main difference is that these AI technologies are focused on doing a small range of tasks really well. Ask your Apple Maps application to write you a speech for your upcoming event, and you probably won't get the result you're looking for. For that, the class of 2023 AI technologies has you covered though.

The Shift to Broad AI Applications

The evolution of AI from specialized tasks to a broader range of capabilities marks a significant shift in the technology's potential impact. While earlier AI systems excelled with specific functions like content recommendation or navigation, the emergence of more advanced AI models will usher in a new era in AI's applicability and functionality.

The initial wave of AI applications was designed for narrow tasks – executing specific functions based on set parameters and data inputs. However, recent developments have seen a transition towards more generalized AI models. These advanced systems, such as ChatGPT-4 or Bard, are not confined to completing a single type of task rather they are capable of understanding and performing a wide array of tasks, from creative writing to complex problem-solving.

At the heart of this transformation are large language models (LLMs). Unlike their predecessors, LLMs use vast datasets and sophisticated algorithms to process and generate human-like text. This enables them to handle diverse prompts, understand context, and provide responses that are not only accurate but also contextually relevant.

Chatbot interfaces have made interacting with LLMs accessible to millions around the world, with the quality of their outputs rapidly improving over the course of the last year. Because LLMs use large data sets alongside complex algorithms, they are able to interpret what we are asking them to do much more accurately and in much more nuanced ways. This has allowed the latest generation of LLMs to support a much broader scope of tasks compared to their predecessors allowing these AI tools to respond with more nuanced text answers as well as outputs in other formats like pictures and video.

The Latest Developments: QSTAR and the Path to AGI

If you have been following the AI space over the last couple of months, you probably heard about the very public removal of Sam Altman as the lead of Open AI. That saga has since played out and Sam Altman has returned to Open AI, but there are lingering questions about why he was ousted in the first place. Some rumors suggest it was because he is working on other projects that are in conflict with Open AI, but others believe it has something to do with some radical advancements that moved the needle a bit too far.

It's important to note that the mission of Open AI and many other AI development organizations is to develop AGI or artificial general intelligence. AGI are essentially human-level capable machines that can think, reason, etc. across many domains. This level of intelligence naturally carries a lot of concerns, and so moving work on AGI forward is important, but not as important as the guardrails we put around for safety.

Speculation has centered around what is called Q*. Q star is a method for training AI to solve problems using increasingly complex reasoning. The idea is that if we can train AI models to solve well-defined problems like those we might find in grade school math, we can train AI to essentially solve problems it does not yet know the answer to. The AI can be trained to reason like a human does when solving these problems, but then can also be trained by playing against itself to solve problems. This shift gives freedom for the AI to find new ways to solve problems it doesn't know. The roadmap is that if we can have AI learn through play, that essentially this method could be scaled up to solve increasingly complex problems.

While it may sound like this potential is pretty far out on the horizon, it is worth noting that there are already antecedents that can serve to expedite this reality. One of the most notable of these roadmaps is that which was used to train AlphaGo the AI that was used to compete against humans in the game Go. Trained by DeepMind, AlphaGo was in partiality trained by competing against itself, which allowed it to learn to adapt its strategies for beating its opponent based on current game conditions. Subsequent versions of AlphaGo have only gotten better growing on what looks like an exponential curve.

As we continue to discover and explore new training methods, find new ways to compress massive amounts of data, and improve computing power, the timeline to more powerful AI and possible AGI will shrink quickly. While diverse opinions exist on the timeline for achieving AGI, the consensus is clear: we are on a path of rapid and exponential growth in AI capabilities.

The Shift to Applications

The last year has been all about LLMs. While ChatGPT appears to be the frontrunner in this regard with their flagship product ChatGPT 4, the rest of the field appears to be catching up. Some projects like Mistral have chosen to go down the path of open-sourcing their models creating a whole new set of opportunities for development on top of powerful LLMs. In time developers will have the opportunity to gain access to the most powerful AI models and build really interesting products on top of them. In short, LLMs will have a substantial impact on new technology, but much of the impact will probably look different than the chatbots that we typically use to interface with these AI models today.

A helpful reference point can be seen on the web. In the early stages of the web, building any kind of web presence was complicated and required a lot of know-how to even build a basic site. Fast forward 25 years and platforms, and low code/no code options mean just about anyone can create a website. In much the same way, it's not hard to imagine that the barrier to building applications on top of AI will greatly diminish with time.

So What? How does this relate to education?

To summarize the story arc of AI in a broad context we can track its growth through an oversimplified yet easy-to-track model of past, present, and next.

  • Wave 1 Past: Narrow AI begins to emerge in consumer applications like recommendation engines, translators, personal assistants, etc.

  • Wave 2 Present: More powerful AI models with much broader capabilities become accessible to consumers.

  • Wave 3 What's Next:

    • New training methods will enhance the capability of AI models to not only produce better outputs but may enable AI to synthesize new information.

    • Applications built on top of AI will enable us to leverage powerful AI models in novel ways.

If we overlay this arc on education, we can begin to forecast the potential application of AI in education past, present, and future.

Wave 1: Education's AI Past

Hindsight is always 20/20, so not much forecasting here. AI's education is primarily compressed with most of the applications of AI occurring within the last decade. Dr. Philippa Hardman captures these roots highlighting areas like automated grading, personalized recommendation engines, and tutoring systems. Much like the examples of AI that we have become accustomed to in everyday use (Alexa, Siri, Google Maps, etc.) focused AI can assist learners or instructors in tasks like grading writing, or recommending a personalized selection of resources based on learner needs. The latter of which is a key feature in many benchmarking or testing platforms and has been in use for some time. The emergence of more powerful AI models will only help to further enhance the capabilities of AI that exist in the platforms we use today.

Wave 2: AI in Education Today

The emergence of LLMs has meant that a whole new realm of applications has now been made available to learners, educators, and school leaders alike. The broad capability of LLMs along with the development of multi-modal (text, image, video, audio inputs) interfaces means we can use LLMs in a variety of ways.

A helpful framework for thinking about how AI can help educators, school leaders, etc. is through the tasks that make up our jobs. If our jobs are a compilation of tasks, finding the right tasks for which AI is well suited is an effective way to think about ways AI can help us. Tasks that are:

  • Repetitive and data-intensive

  • Involve predictive analysis

  • Problem-solving and decision support

  • Content development

Just to name a few.

One area that I have found exceptional use in my roles both in schools and within Ed3 DAO has been using AI as a thought partner. My good friend and Co-Founder at Ed3 DAO, Vriti Saraf, captures this use case brilliantly in her article AI The Cognitive Friend We've Always Wanted. In the article she details how chatbots like ChatGPT can be our cognitive sidekicks, helping us brainstorm, develop, and plan. What I find most compelling about this use case is that it can used nearly universally across roles in the realm of education.

To round out this section I wanted to include some other examples of how AI is making waves in education today. This blog post highlights a list of 43 examples of where we are or might see AI appear in education. Just to sample this list some of the more interesting examples include:

  • Adaptive learning technologies

  • Facilities management

  • Designing learning experience for universal access

  • Data and Learning Analytics

A gentle reminder here; the AI models and tools we know today are the worst that they will ever be. They will only become more powerful and better at executing the tasks for which they are designed.

Wave 3: Transformative Implications for Education and Beyond

Now onto the hardest part, forecasting the future. While I don't have a crystal ball, we can begin to extrapolate how the advances noted above might change education. The creation of new applications built on top of AI along with new training methods will undoubtedly expand the possible use cases for AI and most likely become increasingly accessible with time. I would contend that we will see growth across all compute platforms with the emergence of unique platforms that can build off of new interfaces. Humane is an AI startup building a mobile device in the form of a pin that carries the capabilities of a modern smartphone. Expect more experimentation as AI affords us the opportunity to leverage new interfaces including pins and potentially ushering in a new era of augmented or virtual reality. New computing platforms will enhance agency and access to not just the web but intelligent hardware; a true paradigm shift.

New applications and training methodologies will usher in a new era of educational opportunity. The convergence of these two innovations may lead to innovations that are both sustaining (enhance current innovations) and or disruptive (shifting from the traditional paradigm). Let take a look at some possibilities:

  1. AI as a Generator of Novel Educational Content: Future AI models, equipped with advanced training methodologies, won't just analyze existing information but will synthesize new educational content. Imagine AI systems that can create original educational resources – textbooks, scientific research, historical analyses – tailored to the curriculum and learning objectives. This shift from curation to creation could redefine the role of AI from a facilitator to an originator in the educational process.

  2. Personalized and Adaptive Learning at Scale: Building upon the current trends in personalized learning, future AI will take this to a new level. It will adapt not only to a learner's learning style and pace but also to their emotional and psychological state, offering a truly holistic educational experience. AI could dynamically modify lesson plans and teaching strategies in real-time, ensuring optimal learning outcomes for each learner.

  3. AI-Driven Predictive Analytics in Education Management: AI will likely play a crucial role in predictive analytics, aiding educational institutions in decision-making processes. From predicting enrollment trends to identifying potential dropout risks, AI's predictive capabilities will enable more proactive and strategic management in education.

  4. AI as a Collaborative Learning Partner: Future AI systems could function as collaborative partners in the learning process. These AI 'mentors' could engage in complex discussions, assist in problem-solving, and provide guidance on projects, essentially serving as an additional, highly knowledgeable member of the learning community.

  5. Immersive and Interactive Learning Environments: Leveraging advancements in AI, virtual and augmented reality could create immersive learning environments. Learners could explore ancient civilizations, dissect virtual organisms, or simulate complex scientific experiments in a safe, controlled, and highly interactive AI-enhanced virtual space.

This next wave of AI technology will not just be an extension of its AI's current trajectory; it represents a paradigm shift. This wave has the potential to democratize education, personalize learning experiences at an unprecedented scale, and upend traditional methodologies and systems of learning.

Wrapping up:

Through this quick walkthrough of the history of AI, from its early days in narrow applications to the capabilities of large language models, we've seen how AI is poised to transform education. From everyday tools like translation services and navigation systems to advanced models capable of creative problem-solving and nuanced interactions, AI is reshaping the educational landscape. The anticipation of future advancements, especially in the realm of AGI, suggests a paradigm shift not only in technology but also in how we learn, teach, and manage educational systems.

As we stand on the brink of this new era, our challenge is to navigate this evolution responsibly. The future of AI in education promises a more personalized, immersive, and accessible learning experience. However, it also calls for a commitment to ethical use, equitable access, and a focus on enhancing human potential. As educators and learners, our task is to guide AI's integration thoughtfully, ensuring that it serves to empower and uplift everyone in the educational community.

27 views0 comments


bottom of page