Hello there!
I am Ankit Gupta, a PhD candidate in Computer Science and Engineering at Michigan State University (MSU). I am also enrolled in MSU's Dual PhD program in Ecology, Evolution, and Behavior (EEB). I started my PhD in Fall 2023 and work with Prof. Emily Dolson at the ECODE Lab, which is part of the BEACON Research Group at MSU.
Before joining MSU, I earned my undergraduate engineering degree in 2021 at the Indian Institute of Technology, Kharagpur (IIT-KGP), India, spending five of the most precious and memorable years of my life there. I then pursued MS in Cognitive Neuroscience on a full scholarship in Taipei City, working with Prof. Philip Tseng and graduating in May 2023.
Research Interests
I am interested in exploring fundamental challenges in Machine Learning and AI using Evolutionary Algorithms as a tool for optimization, model discovery, and understanding the vulnerabilities of learning systems. I am also interested in synthesizing and studying open-ended AI systems while exploring the deeper questions of how and why they develop certain capabilities. I am deeply captivated by how evolution, as an algorithm, can drive emergent complexity in both natural and AI systems. Specifically, I draw inspiration from the open-ended nature of natural evolution, which, in a single ongoing process, has generated all life forms on Earth, leading to unprecedented complexity and the human-level intelligence we strive to replicate in AI systems. I believe this open-ended framework is key to developing AI systems capable of artificial general intelligence (AGI) and beyond.
As part of my PhD candidacy exam, I submitted a report and presented to my qualification committee, synthesizing three key studies on Exploring open-endedness in Silico—Paired Open-Ended Trailblazer (POET) (Wang et al., 2019), Emergent Tool Use from Multi-Agent Autocurricula (Baker et al., 2020), and MAESTRO: Open-Ended Environment Design for Multi-Agent Reinforcement Learning (Samvelyan et al., 2023). This allowed me to engage deeply with open-ended AI, a topic that continues to shape my broader research perspective.
Research Journey
Before PhD
My research journey has been shaped by a diverse set of research experiences, all grounded in programming challenges at the core of addressing questions I found intriguing and the opportunities I embraced. It began with a summer research internship in Taipei during my third year of undergrad studies, where I worked on an experiment investigating human unconscious perception and its potential role in biased decision-making. The following summer, during an internship at Harvard, I worked on an agent-based simulation studying evolutionary origins of altruism. For my thesis project at IIT, I worked on a simulation platform incorporating a multi-agent system of Braitenberg Vehicles, leading to emergent swarm behaviors under different wiring configurations. In the summer between graduating from IIT and starting my MS in Taipei—and again the following summer—I developed an agent-based simulation platform, exploring potential evolutionary origins of language as part of the Google Summer of Code at an open-source organisation named Red Hen Lab.
Studying the fundamentals of neuroscience during my MS deepened my curiosity about the why questions, as I felt fascinated by the intricate neuronal mechanisms making me ponder how random chance encounters could lead to something as complex as the human brain. This inspired me to work on AI systems emulating or surpassing human-level intelligence using an evolutionary framework, leading me to MSU, where I found a thriving research community and a large, collaborative group deeply engaged in digital evolution research.
First Year
As I began my PhD with an interest in applying evolutionary algorithms to machine intelligence, I started with a simple class project evolving wiring configurations in Braitenberg Vehicles selected for desired agent behavior. This provided an initial hands-on introduction to evolving agent behaviors. Towards the end of that semester, I attended an MSU talk by a Google DeepMind researcher on model discovery from scratch (AutoML-Zero), which sparked my interest in evolution-inspired machine learning. Wanting to further explore this direction, I spent the following summer at an AI research lab at Cedars-Sinai in LA, working on TPOT2, an AutoML tool optimizing ML pipelines through genetic programming.
Second Year (ongoing)
After this, through another class project, I became interested in the vulnerabilities of machine learning models, particularly in the context of adversarial attacks. This led me to investigate how evolutionary algorithms could be applied to study and generate adversarial examples. I have since been working on understanding the foolability of machine learning classifiers using adversarial attacks generated through a simple evolutionary algorithm. This framework evolves adversarial images and evaluates classifier robustness based on model confidence and perceptual similarity. Working on this problem has motivated me to pursue a deeper study of the root causes of model vulnerabilities in learning systems, which I now aim to explore further.
For more details please check my Google Scholar / Publication Summary Page / CV. I have been doing research so far in parallel with fulfilling my PhD coursework requirements. I am on track to complete ten 3-credit classes by Spring 2025 and look forward to devoting more time to research thereafter.
I am grateful to these institutions which have been an integral part in my research journey.
Also, I am grateful for the guidance and support of people in the pictures below.
My labmates and my current advisor:
My previous advisors, mentors, supervisors, and course instructors during my research journey until starting my current doctoral degree:
