Back in 2002, I was using Neural Networks to classify the physical parameters of galactic halo stars in grad school. I was impressed by neural networks because they were so FAST.  My software could estimate 6 physical parameters of 1,000,000 stars every 10 minutes.  Fast forward 10 years after this work, and I learned about using neural networks for image recognition when the AlexNet paper was published in 2012 by three researchers from the University of Toronto.  AI could now do a great job of recognizing thousands of types of objects in images of those objects.  Fast forward another 10 years and those same techniques from AlexNet were then generalized to create ChatGPT.  The YouTube video from Welch Labs I’ve included here does a great job of describing the key concepts in AlexNet for recognizing images, and how these same concepts allow ChatGPT and other Large Language Models to do their work of answering questions for users.  Querium is now leveraging the power of LLMs to augment our patented StepWise AI math tutor to bring the benefits of Generative AI to math students, while avoiding having LLMs hallucinate and generate incorrect answers for students.  We’ll be telling you more about our work enhancing StepWise later this year.  Stay tuned.

If you’d like to keep up with what Querium is doing in Generative AI, drop us a note at ‘smarter@querium.com‘, and we’ll send you occasional updates by email.  If you’d like to use our consulting services to help create innovative AI solutions for your business, we’ll be happy to talk about applications outside of education as well.

Kent Fuka

Founder and CEO