As a senior undergraduate student studying artificial intelligence, I’m constantly amazed by the transformative power of AI in our world. However, with the exciting progress of deep learning, our understanding of intelligence still remains limited.

First, how to describe the computations happening in our brain? How do these computational principles give rise to 1) the various findings of neural data in neuroscience research, 2) our behaviors, like learning and memory, and 3) our consciousness of self and pursuit of self-realization? Sometimes, however, it is more convenient to ask the inverse question, i.e., with the findings of the experimentalist, what model should we use to best describe these results? What do these models tell us about the underlying mechanisms of our brain?

Second, how do we understand the gap between biological intelligence and the current machine intelligence? Is the current deep learning paradigm capable enough to build highly intelligent agents like humans, so that the differences between ANN and our brain are merely implementation details, which are caused by different computational architectures and optimization algorithms? Or, are some biological features, like ongoing plasticity and modularity, essential for next-gen machine intelligence? In that case, what can we learn from the cognitive features and neural architectures of humans to build better AI?

Driven by these puzzles, I’m currently working at the intersection field of neuroscience and AI.