It's been over a year since Albert shared his vision for 'tutors for everyone,' and we're all the more convinced that AI will not only make knowledge more accessible, but also unlock radically new learning experiences.
The sad truth is that our institutional education system remains fundamentally broken and unable to keep pace with the technological changes happening all around us. This isn’t anyone’s fault per se, but simply the very nature of institutions being better suited for incremental change rather than radical transformation.
Meanwhile, in contrast to this institutional inertia, we’re seeing a rapid acceleration of innovation in AI learning. In just the last year, we’ve seen the emergence of homework helpers, essay writers, language avatars, and even tools that quite literally help students cheat on exams. It’s enough to make you question whether there’s even a point to learning when we have so much technology at our fingertips (a subject for another blog post). But putting that aside, we think of this first wave of AI learning as showing us that AI can continuously produce and tailor the knowledge we need learn effectively.
We’re now entering a second wave where AI is pushing the boundaries of how we learn, regardless of our individual learning styles. Here’s some hunches about what the future of AI-powered learning might look like:
#1 POV on personalization: It's well understood that people have diverse learning preferences, and that tailoring teaching methods to individual learners can be hugely valuable. However, translating these insights into product is challenging. We're excited about approaches with a strong POV on how they want to personalize their UX and, as a result, can quickly dial into the user's unique learning styles and adapt in realtime. Duolingo's adaptive difficulty is a good example. Their algorithm adjusts exercise difficulty based on user performance, keeping learners in the "zone of proximal development" - challenging enough to learn, but not discouraging.
#2 New form factors: The platforms that thrive will be those that create entirely new format factors for learning that just weren't possible before large language models. Imagine a system where humans sense-check their understanding by teaching concepts to agents in a kind of real-time loop. Or environments that seamlessly integrate your physical surroundings, like studying botany on a hike or art history at the Met. Or learning physics from Einstein or creative writing from Jane Austen. These experiences will not simply iterate on existing models, but truly reimagine what learning can look like.
#3 AI-powered learning communities: One of the most interesting possibilities is that AI reshapes learning communities by facilitating co-learning between humans and agents. We’ve seen the impact of peer-to-peer learning. And we’re seeing what happens when you introduce AI to get peer-to-agent learning. This will eventually evolve into networks of humans and agents co-learning together, unlocking new ways of collaboration and knowledge acquisition that are hard to fully picture today.
#4 Unbundling the traditional school: As AI-powered learning rises, the traditional school “package” will continue to come undone. Schools have historically been responsible for far more than just educating students; they’ve provided socialization, affordable childcare, physical activity, nutrition, and (mental) healthcare. We’ll see a proliferation of tech-native point solutions that meet these needs in new and creative ways.
#5 Motivating the unmotivated: Many AI-native learning platforms target users who are already self-motivated to learn. But we also need to consider how to support people who currently rely on institutions like schools or universities to force themselves to absorb information. You can bring a horse to water, but what does accountability look like in an AI-powered world with abundant individualized learning? Could AI companions or peers play a role in keeping learners on track? Can we build systems that nudge learners when they fall behind? These questions are important to address as the old model of institutional discipline breaks down.
If you’re building in AI learning, and / or have thoughts on these hunches, we’d love to hear from you.
