June 2018 – Stanford AI4ALL

Day 4: Classification, Computational Biology, and Hearing from Telle Whitney

Eugene continued his lecture on machine learning this morning, this time with a focus on classification. He explained that binary classifiers can be evaluated using metrics like sensitivity, specificity, and precision. The lecture focused on two main classifiers: decision trees and Naive Bayes. Students created an example of a decision tree that could help them decide whether or not to buy a certain item from a grocery store. To better understand the Naive Bayes classifier, campers thought back to their first lecture with Eugene on probability and statistics and looked at an example of classifying whether or not an email is spam.

Eugene explains how to evaluate a classifier.

Eugene touched on various other supervised learning algorithms such as k-nearest neighbors, logistic regression, and support vector machines, as well as a few unsupervised approaches, such as clustering and principal component analysis.

Campers interact with a visual demo to explore machine learning.

Campers then heard from Stanford Professor Anshul Kundaje, who explained how machine learning is used to identify the genomic causes for certain diseases. While manually sequencing a person’s DNA would be incredibly time-consuming, advances in technology have made it possible to speed up this process immensely, making it much easier to collect data on the human genome. Unsupervised machine learning can then be applied to this data to find variations that may be related to certain diseases.

Prof. Kundaje suggested that artificial intelligence will greatly impact the future of personalized medicine, as individuals will be able to receive personal diagnoses and treatment suggestions based on their genes.

Students go up to Prof. Kundaje after his lecture to ask questions.

After lunch, Wells helped the girls review the machine learning concepts they’d been learning the past few days.

Campers then got into their research groups for the first time. This year, there are four research groups: Computational Biology, Computer Vision, Natural Language Processing (NLP), and Robotics. After introductions with their mentors, the students were given an overview of their research fields and the projects they’d be working on for the next two weeks.

The computational biology group asks about neural networks.

NLP mentors give an overview of their research project.

Campers then had the special opportunity to hear from AI4ALL board member, Dr. Telle Whitney. Dr. Whitney told the girls her story of how she became a co-founder of the Grace Hopper Celebration of Women in Computing and CEO of the Anita Borg Institute.

Campers were inspired to hear how Dr. Whitney struggled with imposter syndrome but overcame it, and they were left with the advice to make connections, take risks, and surround themselves with supportive mentors and friends.

Campers were incredibly inspired by and grateful towards Dr. Whitney.

We hope the campers have had a fun first week and are enjoying all they’ve learned so far!

Blog post and all photos by Anna Wong.

Day 3: Machine Learning, Refugee Integration, and Confidence

Today, Eugene began teaching the campers about machine learning. After discussing what machine learning is and why it’s important, the students learned about the differences between various types of machine learning, including supervised, unsupervised, and semi-supervised learning.

Campers learned about the purpose of linear and polynomial regressions, and the importance of neither underfitting nor overfitting the data. Eugene taught the students how to evaluate the accuracy of a model with cost functions and cross validation, as well as how to use stochastic gradient descent to minimize the error of a model.

Campers discuss potential inputs for a regression.

Following Eugene’s lecture, the students heard from Jens Hainmueller, a professor in the Department of Political Science and the Business School at Stanford. Prof. Hainmueller is part of the Immigration Policy Lab, which works with governments to address the current refugee crisis. After giving some background information on the refugee crisis, Prof. Hainmueller described how artificial intelligence and data mining are being used to geographically assign refugees and improve the integration of refugees. He and his lab use supervised machine learning on historical data to determine where each refugee should be placed in order to maximize their probability of employment.

In the afternoon, campers had time to discuss the machine learning concepts they’d learned earlier in the day. Morgan Ames, an ethnographer for AI4ALL with a background in computer science, worked alongside Wells to help the students review.

Wells helps campers understand weights and error.

Morgan explains how weights of a model are updated.

The campers then got to hear from two women who work at Google, Sherol Chen and Melanie Warrick. Sherol and Melanie told the girls how they became interested in technology and how they got to where they are today.

They explained the importance of soft skills such as confidence and resilience, and encouraged the girls to advocate for both themselves and for others. Sherol and Melanie stayed afterwards to continue answering the many questions that the girls had.

As we near the end of our first week, we hope the campers are excited for all that lies ahead!

Blog post and all photos by Anna Wong.

Day 2: Graph Searches, Computational Linguistics, and Jackrabbot

The day began with another lecture by Eugene. After going over the previous day’s challenge problem, which many of the students had come up with solutions for, Eugene taught the campers about graphs and graph traversal algorithms.

Campers learned about various graph search algorithms, including breadth first search, depth first search, and Dijkstra’s algorithm. They spent some time exploring interactive online demos of the searches, then discussed the pros and cons of the algorithms as a class. The campers learned how these algorithms can be implemented in computer programs using data structures such as stacks, queues, and priority queues.

Campers work together to demonstrate how the depth first search algorithm works.

Next, campers got to hear from Stanford Professor of Linguistics and Computer Science Dan Jurafsky about his work in natural language processing (NLP) and computational linguistics. Prof. Jurafsky started with a fun analogy about the origin of national foods, eliciting lots of laughter from the campers.

He then moved on to describing how NLP is used to research issues that are very relevant today. For example, classifiers have been utilized to analyze traffic stop interactions and measure how polite and respectful police officers are to drivers of different races. Campers also learned how NLP and computational linguistics can be used to explain the differences between the language used in negative and positive restaurant reviews.

At the end of the talk, campers were full of questions about this research. Luckily, they got to spend their lunch with Prof. Jurafsky to continue learning about NLP.

After lunch, the campers spent some time reviewing probability and graph searches by working together on a challenging problem set, with the help of Wells.

The girls then got to meet the Stanford Jackrabbot, an autonomous robot that uses many cameras and sensors to navigate around. Marynel Vázquez, one of the researchers who has worked on Jackrabbot, demoed the robot while explaining how it works.

The campers had tons of questions about how Jackrabbot portrays different expressions with its eyes and how researchers are making Jackrabbot safe.

After dinner, the students worked on their presentation skills with Alex Dainis, who is a part of the Oral Communication Program at the Stanford Hume Center.

They learned how to craft an engaging introduction and how to use visuals to enhance their presentations. The girls all took part in an elevator pitch activity to practice their delivery skills. They came up with short pitches about AI4ALL, projects they’ve been working on, and more.

Students work on crafting elevator pitches.

A camper delivers her pitch to a partner.

Blog post and all photos by Anna Wong.

Day 1: Statistics, Computer Vision, and Prototyping

This morning, campers dived right into learning statistics and probability with a lecture by Eugene Davydov, a software engineer at Google. He introduced the concepts of probability and counting to students, defining terms such as sample space, independence, and random variables, and he used many examples, including the famous Monty Hall problem, to help students understand the ideas. With Eugene’s encouragement, the girls asked and answered many questions to clarify concepts. Eugene left them with a challenging Bayes’ rule problem to work on later in the day.

Eugene’s lecture was followed by guest speaker Olga Russakovsky, a professor at Princeton University and one of the founding members of AI4ALL. Prof. Russakovsky talked about computer vision, giving an overview of the standard computer vision algorithm.

Students learned about the many applications of computer vision, from autonomous cars to medical assistants. Prof. Russakovsky pointed to many of computer vision’s successes, such as sorting mail, depositing checks, and detecting faces, but also acknowledged the challenge of data imbalances and biases. She emphasized the importance of diversity in computer vision and artificial intelligence, explaining that people who come from different backgrounds are able to help identify and mitigate the data imbalances present in artificial intelligence.

At the end of Prof. Russakovsky’s talk, campers were excited to ask insightful questions, many of which revolved around ethics, fairness, and bias in artificial intelligence. The campers then got to continue asking Prof. Russakovsky questions as they enjoyed lunch together.

IMG_6927-edit — The campers had many questions for Prof. Russakovsky.

After lunch, the girls had time for independent work, where they reviewed the morning lecture on probability and statistics, worked on solving the challenge they’d been given by Eugene earlier in the day, and practiced coding with Python challenges.

Wells Lucas Santo, education manager at AI4ALL, helped students review, answering their questions and giving them an additional probability proof problem.

Wells explains how to solve the additional challenge he gave the campers.

After spending their time reviewing, the campers visited the Volkswagen Automotive Innovation Lab (VAIL), which is part of the Center for Automotive Research at Stanford (CARS). They met Stephanie Balters, a postdoctoral fellow at Stanford, who described how the lab incorporated a form of therapy into the car, using a massage chair to guide the driver’s breathing and help the driver relax and calm down.

She next guided the campers through a prototyping activity. Campers were first asked to brainstorm negative emotions and the physical reactions associated with those emotions.

Then, they came up with ideas for products that could detect these reactions and built prototypes of pulse monitors, pressure sensors, and salt detectors (to detect tears), presenting them to the group.

The campers’ first full day was filled with learning, and we can’t wait to see what’s next!

Blog post and all photos by Anna Wong.

Day 0: Arrival and Welcome Dinner!

Today, Stanford AI4ALL, formerly known as SAILORS, welcomed 30 new campers into their dorms in Synergy house. After settling in, the campers and their families went outside to start getting to know one another before dinner.

Program director Rick Sommer kicked off the welcome dinner by introducing the AI4ALL program, discussing the program’s history and its evolution from SAILORS to AI4ALL.

Dr. Juan Carlos Niebles, a senior research scientist at the Stanford AI Lab, then briefly introduced everyone to the four research projects that campers will be working on throughout the program.