Taking a Deep Dive into AI
Guest Post by Irene Yang, Stanford AI4ALL ‘17
AI4ALL Editor’s Note: Irene Yang participated in AI4ALL’s first AI Research Project Fellowship program, where she was paired with a mentor who works in AI to collaborate on an AI research project over the course of 3 months. She worked with another mentee, Janice Yang, to identify the severity of diabetic retinopathy using computer vision. Below she shares her personal learnings and her experience at the final fellowship meeting. Learn more about AI4ALL’s Research Project Fellowship Program and other mentee projects here.
Three months ago, I would never have imagined that I would complete a computer vision project and learn so much about AI.
The start of AI4ALL’s Research Project Fellowship Program was a little daunting because the work ahead of us seemed so long and complicated. However, with the help of our amazing mentor, Andrea Frome, my teammate Janice and I completed our project on identifying the severity of diabetic retinopathy and presented our work at the final meeting of the program.
Throughout the program, I learned so much about AI from a technical perspective and gained new insights about how to use AI. While working on the project, I could not help but think back to my Stanford AI4ALL experience. It laid the foundations for this mentorship project, which gave me a chance to delve into the higher-level concepts.
At Stanford AI4ALL, datasets and the data pre-processing procedure were all provided for us. On the other hand, for the mentorship project, we had to figure out all this on our own. I discovered that data is not usually clean and balanced. As a result, we had to come up with strategies to reorganize the data we found. Also, during the AI Research Project Fellowship program, we explored several techniques to pre-process images, while at Stanford AI4ALL, we only used one technique.
Overall, there was much more exploration in the mentorship project, which I enjoyed. We encountered a diversity of errors throughout the course of our project, and each was a learning experience.
Andrea, our mentor, was a great help to us as she offered possible solutions to the problems and several data analysis procedures. We tried out various solutions and methods to analyze the results and determine which method worked best. I was glad that I was able to experience the experimental process with the help of our supportive mentor.
On the day of the final presentations, Professor Fei-Fei Li talked about AI in the future and discussed three main points about human-centered AI. First, she emphasized the need for a deeper understanding of neuroscience and cognitive human behavior so that AI can understand intentions and become a broader field. Professor Li also argued that AI should be used to augment and enhance the jobs of humans, not to replace them. For example, disaster relief can be improved with AI by quickly identifying calls for help, and where they are coming from. Professor Li’s final point was that in order to make AI less biased, the field needs economists, legislators, philosophers, and professionals from other disciplines. Including multiple perspectives will make AI a more fair and useful tool.
I realized that I could connect these points — about AI striving to be human-centered, an augmentation to human work, and created by a broad and diverse group of people — to our project. Our final algorithm was far from perfect, with an accuracy of 64%. I learned that AI has its limits — it can address some of society’s current issues, but it isn’t a panacea.
When we began the project, we did some research on DR and found that DR’s diagnosis process is often tedious and not always very accurate. AI probably is not ready to take over the task of diagnosis, but Janice and I thought that AI could be used to enhance and improve this task.
I gained valuable experience from the AI Research Project Fellowship, not just on a technical level, but also through the process of building a project from scratch.
About Irene
Irene Yang is a sophomore at Dougherty Valley High School. She attended the 2017 Stanford AI4ALL program, which sparked her interest in AI. She loves being involved in the STEM field, so she volunteers at the Chabot Space and Science Center, where she teaches science concepts through interactive activities for visitors. She also participates in STEM clubs such as Girls Who Code and WiStem. Irene is also very passionate about music. She has won several awards at piano competitions, and she is a member of her school’s Wind Ensemble and Band Council.
