This is an ongoing reflection I'll keep adding to as the semester continues.

This semester I am enrolled in a class called Humans, Animals, and AI with Dr. Justin Wood. This is already one of my favorite classes I have taken here at IU Bloomington. I have learned so much about ideas about how to model cognition, how to implement learning, and what parts of our brains are innate versus learned.

In the past, AI researchers were hardcoding knowledge into their models. If conditions were met, a specific answer would be given as output. As we have evolved in the realm of AI, DNNs, reinforcement learning, and modern techniques, I have learned that there is so much more that an AI model has the capacity to do.

You always hear people say, "It's okay to mess up, as long as you learn from your mistakes," and I think AI is taking on a bit of the same framework.

As a newborn, we have the ability to perceive our spatial world, create maps, and sense hierarchy. This is also the general structure of a modern AI model.

Another thing we learn about in class is that as humans, we are built with a rigid body and a flexible brain. Species like elephants have an incredibly flexible brain as well, but their rigid body is somewhat limiting in what they are able to do. Evolution has shaped us into a body that is perfect for interacting with our spatial world, but our mind is flexible enough to learn and adapt to whatever we encounter within it.

We then process data through the experience of the world. Through all of the different systems in our brain, we constantly process and feed our brains data, and our brain develops other parts specific to our environment.

We learn what faces are, what objects are, what is alive versus not, and the idea of object permanence. There is so much that we develop in our brain just from the input of data through our environment. This has been so cool to learn about, and how we have now shifted to program AI the same way.

We feed AI endless amounts of data. AI systems are no longer specialized to just play chess or solve math problems. They can now have conversations about feelings, edit images, receive video inputs, and write essays, all in one model. They learn to predict, and if they predict incorrectly, they adjust and try not to make that mistake again.

It is so fun to learn about human cognition and what happens when changes are made to our environment. So much of our world is just about prediction, and this is exactly what an AI model is trained to do. It is interesting how the more we learn about the human brain and its limitations and expertise, the more we also learn about AI and how it functions. This shows how interconnected we are to these new AI models, which is part of why they feel so intuitive to interact with. In a way, they learn similarly to us, just without a body, emotions, or lived experience.

This class has made me significantly more excited about the future of AI and the possibilities, especially seeing how closely modern models mirror the way our own brains build knowledge from experience. I want to continue to learn and grow on this knowledge and hopefully be an asset to technology teams and working on training and creating highly specialized models in the future.