The Center for Data Innovation spoke with Bodo Hoenen, U.S. CEO of NOLEJ AI, a Paris-based startup that uses generative AI to create interactive content for the classroom. Hoenen spoke about how generative AI can dramatically improve the rate of learning and implementation of new skills.
Gillian Diebold: What are the benefits of generative AI to the classroom?
Bodo Hoenen: Firstly, I want to provide some context of what the challenge is. The challenge is that 42 percent of the jobs that our school systems are preparing our kids for will be obsolete by the time they graduate. In addition, if we look at lifelong learning and the market that is generating courses, 82 percent of the course content that they produce is not relevant to the learner’s current needs. This means that by the time they do need to use what they’ve learned, they probably have forgotten what they’ve learned. And the reason for this is that it takes an incredibly long time to generate courses; it takes anything from several months to even several years to build the courses that we use within our public school systems, universities, and online programs. The benefit of generative AI is we can flip that script totally on its head, and we can auto-generate curricula in real time based on any learning goal that students may have. This is a game changer for how we think about education. And it solves those two challenges I’ve just highlighted.
Diebold: How does the NOLEJ product work?
Hoenen: NOLEJ comes from 10 years of research. We’re looking at the challenge from a very holistic sense, asking what learners and educators need in order to be able to solve these types of educational challenges. And we’ve come up with the idea of building three tools. The first is a tool that can rapidly generate microlearning in almost real-time. It’s able to take any suitable information and generate micro, interactive eLearning content. The second is NOLEJ Graph, which maps out these items onto a multi-dimensional latent space. You can think of it as Google Maps for learning. This allows you to query that map and say, “Hey, you know what, I want to learn X or Y or Z,” and we can generate a map of all the concepts that lie between your current understanding and that learning goal in real time. So now, you can generate curriculum in real time. The third and final component we’re building is NOLEJ Protocol, which is a protocol that guides you through this map like a GPS would. And the GPS guides you not only to content but, more importantly, to other learners, teachers, and experts along the way. Most of the learning takes place through interactions with other educators. These three tools we’re building are focused on solving those three challenges. The NOLEJ product is really geared towards allowing educators and learners to generate curriculum and then connect to anybody who can teach them those particular concepts along the way.
Diebold: Can you talk about a successful use case?
Hoenen: This is a personal story. While we were working on solving these challenges, my daughter suddenly became paralyzed. She lost the use of her arms and is almost completely paralyzed. But we knew through Google searching that there were some potential solutions to this. The doctors told us that there was less than a two percent chance of recovery. But we knew that there were some types of solutions out there. For example, there were brain-controlled exoskeletons being designed behind closed doors at famous universities. They wouldn’t share these plans with us. So we thought, well, could we design one ourselves?
It took us several months. Using NOLEJ AI, we deconstructed everything that we could find about exoskeletons, brain-machine interfaces, 3D printing, about signal processing. We mapped it all out into NOLEJ Graph, which provided us with a map of our ignorance. From there, we started to go through that map using NOLEJ Protocol and were put in touch with experts all around the world. Within seven months, we built what our doctors thought was impossible. In just seven months, we learned what we needed to learn about brain-controlled exoskeletons, plus we actually built one which allowed my daughter to use her limbs again. This is a great use case, but it also highlights something different from if we had gone the traditional route. To build a brain-controlled exoskeleton on the traditional path, well, first, we’d need to go through several years of medical school, and then we’d need to specialize, and then maybe seven years later, we can start building this exoskeleton. Well, we did it in seven months. This is a profound difference in the speed and ability to learn, where you’re focusing on the learner’s intrinsic motivation, as opposed to providing them and force-feeding them a linear course.
Diebold: How do you safeguard against the potential risks of AI in education?
Hoenen: One of the big risks at the moment is trust. AI can hallucinate really well; it can hallucinate very confidently. How do we address this? Well, the NOLEJ AI product, the way you work with it, is you have to provide it source material. That source material could be a PowerPoint, it could be a PDF document, or a chapter from a textbook, whatever the case may be. You provided some source material. That source material becomes the ground truth, and our AI can’t hallucinate anything that is contrary to that source material. This is a very important guardrail. It puts the teacher in the driver’s seat of using the AI. It ensures that the teacher has control over what type of biases the AI is going to produce because you have control over the source material you provide the AI. So, you can eliminate bias and make sure that you have the correct output being produced. This is critically important when we think about generative AI in education. We need to make sure that the educators are in control and that they are able to direct the generation of content appropriately.
Diebold: What does the future hold for NOLEJ?
Hoenen: I think we’re at a unique point. Most of the industry is looking at how we can use AI to optimize the current educational system. How do we use it to solve the current educational system’s challenges? We’re thinking of it a little differently. We have come at this through 10 years of applied research, working in environments that lack traditional educational infrastructure, like the slums of East Africa or refugee camps, or even helping to teach Afghan girls who aren’t allowed to go to school. Now, how do we solve those challenges?
We started to think about the learner’s needs and the educator’s needs instead of the institutions and educational market. If I look at the future, I think there’s a huge opportunity now for us to reconsider solely trying to optimize the current system. Let’s not just try to optimize it with 10 to 20 percent improvement. Let’s think about a game-changing improvement in the way that we learn and the way that we teach.