AI 4 ALL
Throughout my high school career, I’ve been trying to find ways to learn more about the community that surrounds me. Last year, I started a Girls Who Code chapter at my school to bridge the gap between the exciting tech industry that surrounds us and my high school.
I knew I wanted to dive deeper with this club and have it actually make a difference in my school; many clubs with similar themes had come and gone but were largely just college application driven, free pizza providing weekly meetings. In order to help our GWC club avoid a similar fate, I needed to expand my STEM knowledge so I turned to Google and searched for STEM opportunities I could apply to this summer.
That search led me to a program called AI4ALL, hosted by the wonderful Berkeley Artificial Intelligence Research (BAIR) Group. Aside from seeing Waymo vans jerking to a stop at local red lights, I didn’t know anything about AI and figured this was the perfect opportunity for me to learn.
AI4ALL is an underrepresented minority-focused camp that aims to grow the next generation of technology leaders and thinkers. For 5 sunny days, I commuted daily from my family’s home in Milpitas, CA to UC Berkeley, 40 miles north.
The first day of camp was centered around introducing the students to each other and to the subject matter. Before this, I didn’t know what AI was or how it was used in the real world. By the end of the lecture, I understood that AI doesn’t really have a formal definition and that it has the potential to impact our world tremendously.
Our project for the week was to make a robot run from scratch. Having only done introductory Python courses at my local community college, I’d be lying if I said I wasn’t intimidated. It had been a while since I had even programmed a simple calculator, much less a robot.
Our camp counselors taught us the appropriate Python libraries and functions we’d need to make our robot work and gave us lots of practice problems to help us apply and retain the material. My ears buzzing with excitement for the days that lie ahead, I boarded the BART train back home.
The next morning, we had a guest speaker who talked about the ethical debate that is ever present in AI. She discussed how we need to control biases in AI systems in order to make their behavior fair. For example, an AI system that determines parole in the justice system could easily be biased against people of color, depending on the data that is used to train the model. To use AI effectively, we need to make sure AI systems reflect our values. This was followed by a quick discussion on supervised learning and its pros/cons/use cases.
In the afternoon, we dove into our projects. We started by gaining familiarity with the codebase. The camp counselors gave us templates of three “robots”: a cheetah, an ant, and a walker. Our goal was to make two of the three robots be able to run. We tinkered for a while, changing body shape parameters, refining step sizes, and reading through libraries to figure out which functions had to be called.
Once I realized how to get my robots to run, I started optimizing for distance. I doubled down on the cheetah, figuring that a robotic cheetah must be named after the real animal for some reason. Turns out, I was right!
Today’s lecture was about the various applications of AI. We learned about the many accomplishments of AI today, like making it easier for people of different languages and backgrounds to communicate via Google Translate, help sales teams stay in touch with their customers with automated messaging systems, and give billions of people the gift of physical mobility through Uber and Lyft. I was surprised to hear that AI was present in many of the products I use every day.
The next question on my mind was: how can I actually make some of these apps? Our next lesson centered on training a machine algorithm. There was a lot more math than I had expected, but fortunately, I wrote a bit of Python logic that did the hard computations for me.
My favorite part of camp was our visit to Cruise Automation in the heart of San Francisco. Cruise is a startup that is devoted to making self-driving cars a reality. We spent the afternoon seeing their office and learning about the various cars and experiments that helped optimize their route (ha!) towards making mass market self-driving a commodity.
After the field trip, we came back to a talk on interpretability and imitation learning, which is the idea that a machine can learn from some source of truth ‘expert.’ In the example, we saw an untrained robot learn from a fully trained robot. The experience was meta, to say the least.
We toured some of the research labs on campus to witness their work first-hand. In one of the labs we visited, we saw a robot making a bed and sorting objects. The implications of such an innovation are quite obvious: imagine the positive effect this type of robot could have on the senior care industry!
After the tour, I gave a demo of my own: a neural network based cheetah robot that could run close to 100 meters. We ended the day with a closing ceremony and it was time for me to say goodbye to a new community.
What I learned
- Diversity is critical in creating and informing ethical AI systems. Without a diverse set of training data, it is very easy for AI to become biased and at odds with our ethical principles.
- AI has tremendous potential in every field from healthcare to agriculture. The opportunities are truly endless.
- AI is not the scary, research-only topic I had assumed it was in the past. AI4ALL was an incredibly formative experience since I found other people in my age group who have similar interests, ambitions, and curiosities.
Thanks for a great summer, Cal! Until next summer :)