For Drexel College of Computing & Informatics (CCI) Computer Science PhD candidate, Maryam Daniali, winning Drexel University's Three Minute Thesis (3MT®) People’s Choice Award in spring 2023 was only a recent achievement to add to her impressive success streak.
Daniali won a third-place Scholarship Award for Research Excellence from Shabahang in 2022. She also interned for Bosch Center for Artificial Intelligence in 2022 and Children’s Hospital of Philadelphia (CHOP) Research Institute in 2021, where she penned a newsworthy study that generated notable media mentions.
The accolades don’t stop there. Daniali was awarded the Anitab.org Grace Hopper Student Scholarship in September 2021, as well as the Outstanding Researcher Award in June 2021 and the Outstanding Teaching Assistant Award in May 2020 from Drexel’s own Department of Computer Science. Before starting her studies at Drexel, Daniali attended University of Tehran to pursue bachelor’s and master’s degrees in computer science.
We sat down with Daniali to learn more about her exciting work during her time at Drexel’s CCI and what she enjoys doing when she isn’t laser-focused on her studies.
CCI: What is the main concept behind your PhD thesis, When AI "Sees" It Wrong: Understanding Weakness and Building Robustness in Vision?
Maryam Daniali: The main concept behind it is the weakness of artificial intelligence (AI) in visual recognition tasks and the need to build robustness in AI systems. While AI surpasses human perception in specific vision tasks, it is far from perfect and can be easily fooled by minor changes in images, such as lighting, perspective, or image manipulation, leading to incorrect identifications or diagnoses. The concept emphasizes that humans have a remarkable ability to recognize objects even with partial or distorted visual information, while AI lacks the same capability. In our research we draw insights from neuroscience to create vision systems that analyze visual information over time, leading to improved robustness and accuracy. The main idea is to showcase the limitations of current AI systems and explore ways to make them smarter and more reliable, learning from human perception.
CCI: What has inspired your thesis?
MD: First of all, AI seems to be everywhere. From face detection on our phones to autonomous vehicles on our roads, AI has become an integral part of our daily lives. However, despite its widespread adoption, I believe a lot of us experience its limitations in day-to-day life. Whether it is facial recognition failing to unlock our phones or misidentifying objects in images, these instances showcase the weaknesses of AI in visual recognition tasks.
I was always curious to explore the parallels between human perceptions and AI systems. Humans can understand context, identify objects from different angles, and make sense of complex scenes. With the rapid advancements in AI technology and their applications in various domains, including security and healthcare, I believe it is more important than ever to address the weaknesses of AI in vision.
Whether it is enhancing security measures to prevent false identifications or improving healthcare diagnostics to avoid incorrect diagnoses, building robust AI systems is essential. By bridging the gap between human perception and AI vision, we can contribute to the ongoing advancement of AI and make it more reliable, accurate, and trustworthy in real-world scenarios.
CCI: How has Drexel University prepared you for this investigation?
MD: During my master's and PhD studies at Drexel University, I gained a strong foundation in AI, computer vision, and research methodology, which have been instrumental in my investigations. Drexel’s diverse and interdisciplinary environment has also been immensely beneficial. Through collaborations with scientists from various backgrounds, such as neuroscience, psychology, and healthcare, I had the opportunity to pursue multidisciplinary research. These collaborations allowed me to explore different perspectives and evaluate our techniques in real-world applications, enhancing the robustness and practicality of our work.
I am grateful for the guidance and support of my mentors at Drexel. Drs. Dario Salvucci and Edward Kim have been particularly invaluable in helping me navigate through challenging research projects. Their expertise, guidance and continuous support have shaped my research approach and contributed to the success of my investigations.
CCI: What activities do you like doing outside of your studies?
MD: Photography has become my new hobby. I like to capture and share the world as I see it. Painting is also an old hobby of mine. My thesis has kept me busy, but I cannot wait to pick up the brush and let my imagination run wild again.