Founder’s Five: Lewis Johnson, Alelo
August 23, 2024 Founder's FiveFounder’s Five is a continuing series from Tyton Partners that invites education company founders to shed light on…
Founder’s Five is a continuing series from Tyton Partners that invites education company founders to shed light on their own success and illuminate the landscape for other education entrepreneurs and investors by answering five basic questions.
Alelo was founded in 2005 at the University of Southern California and later spun out as an independent cloud-based e-learning platform that allows students to engage in interactive online conversations with socially intelligent virtual avatars. The company’s avatar-based artificial intelligence platform creates asynchronous lesson plans and provides virtual role-play simulations that combine innovations in software technology, social science, and learning science, enabling users to acquire knowledge, develop new skills, and improve their performance. The products have supported training for 500,000 members of the military, corporate employees, and students attending higher education institutions.
We started our company, Alelo, as a spin-out of the University of Southern California with an idea that was both timely and ahead of its time. The idea was to learn foreign languages and cultures by practicing with computer-generated avatars that speak and understand human language. I got funding at USC from DARPA in the early 2000s to develop the prototype, an immersive game for learning Arabic, in which you had to speak Arabic with the game characters in Arabic to play the game. This addressed an immediate need because military personnel were sent to Iraq and Afghanistan with little language training. And it proved to be hugely successful, both in training speed and effectiveness. Five hundred thousand military trainees eventually learned foreign languages and cultures using Alelo products.
We then looked for opportunities to show that our avatar-based training approach can effectively train various skills and occupations, not just foreign languages, in the military. The XPRIZE Rapid Reskilling competition gave us that opportunity. We showed that we could train people to become community health workers at least twice as fast as conventional methods and place them in well-paying jobs, and we were one of the competition’s winners. This gave us a good entry into the healthcare market, where we continue to expand.
Generative AI has opened new opportunities for us and expanded the range of training problems we can address. Companies like ours will transform how people learn by providing personalized experiential learning at scale. Within the next year or two, we plan to create thousands of avatars that offer personalized learning to tens of millions of learners.
It means connecting them with the people that they need all around the world. We want to go beyond just the Western countries. We want to help everyone in the world get access to world-class education. That would look like the world that Sal Khan describes in his book, The One World Schoolhouse, where everyone uses technology to give custom and scalable access to affordable or free education to all students worldwide when they need it.
I wish I had known what institutional barriers would stand in the way of bringing innovative training solutions to market at scale. For example, medical training programs are accredited by individual states and countries, making it difficult to offer global or even nationwide training solutions.
Some states require that learners complete a minimum number of hours of training, even though our training methods help people learn much faster than conventional methods.
In the past, the most significant barrier to creating AI-powered learning tools was the so-called knowledge acquisition bottleneck. Content authoring and model training were very time-consuming processes. We learned that generative AI can eliminate the knowledge acquisition bottleneck. Instead of authoring content from scratch, we can feed existing content to large language models to create new AI-powered content. This can often reduce authoring time and cost to nearly zero.
Carnegie Learning was an early leader in AI-powered learning, and I continue to admire its work. I also admire the work that Khan Academy is doing. Squirrel AI Learning is also a company to watch. They have achieved a very large scale and have significant learner data to work with.