Nidhi Parthasarathy, Wednesday – June 29th, 2022
Day 3 began with a camp-wide lecture from another research mentor introducing us to Python. He began with a brief history of computation from the Abacus in 3000 BC, to the Babbage and Turing machines, and the first computer, ENIAC. We also learnt about Ada Lovelace, the first programmer. We then went through the basics of Python — variables, operators, lists, dictionaries, control flow, etc. The next session went into additional details with our smaller cohort including programming exercises.
We went through several code labs. For any student interested in a good overview of Python, a good reference is the “Intro to python” class that I teach at robotics4all. (It is free, so feel free to join!)
The afternoon session featured a talk from Chelsea Finn, professor at Stanford, talking about understanding intelligence through robots.
She started with an introduction to her journey into CS (first lego league robotics, MIT CS, PhD at Berkeley BAIR, and professor at Stanford).
She then gave an overview of the societal impact from robotics, from surgery to search and rescue and other dangerous tasks that humans could otherwise not do. She highlighted how they started with reinforcement learning to teach robots to learn in unfamiliar environments and how their work taught them the importance of generalization and diversity of experiences and data for training. Their other challenge was around having robots effectively predicting the future and fitting the model behavior to demo data using math libraries like Tensorflow and Pytorch.
Her video demos were really cool, and it was great to see the specific application of AI in robotics visually. Robots folding laundry were my favorite. I did not realize how many actions go into something so simple that we take for granted, and I would really love a robot to help with my laundry! 🙂
The next session was a really cool “Smart Summon” AI demo/lecture from Lauren Yang from Tesla. She started with an introduction to what Tesla was trying to do — transition the world to sustainable energy through smarter and safer electric cars. She gave an overview of their cars and the different kinds of sensors that make their cars “smarter.” She talked about their “autopilot” and the different types of self-driving cars from L2 to L5 (user-monitored to fully self-driving). An interesting statistic was the more than 200,000 AI-driven lane changes by Tesla cars every day.
She then talked about the “summon” button in Tesla cars where the cars come to you when called and showed many demo videos (it was so cool to see a driverless car navigating complex roads and parking lots to come to the user!). She went over four key aspects: (1) collecting data (both highways and also on parking lots and driveways), (2) detecting obstacles (speed bumps, pedestrians, geese, etc; data is obtained from both the Tesla data engine database and from real-time sensors), (3) understanding the environment (e..g, predicting people’s movements), and (4) planning and control (planning routes around obstacles, etc).
She concluded with some of the opportunities ahead (e.g., intersections and stop lights). I really enjoyed this talk since it was so futuristic and amazing, but at the same time, it also made me reflect on the responsibility we have in making AI trustworthy to drive around without hitting people. It was interesting to hear about how the data from their existing users was a big advantage to them compared to Google.
As always, we ended with a one hour social. We transitioned into games this time (scribbl.io) and I met a few new people but also spent more time with friends from the previous two days. We all loved the Tesla demo and talked about it a bit between games as well!
Continue reading for day 4.