Describe some of the technologies mentioned in the article that enable self-driving.

Read: Self Driving.pdf
1. Read the above article about self driving. Based on what the article says, in your own words, what are the arguments for cooperation between the vendors developing self-driving technology?
2. Describe some of the technologies mentioned in the article that enable self-driving.

Watch: https://www.youtube.com/watch?v=hx7BXih7zx8
3. For the remaining questions, watch the video from Tesla’s head of AI. This video is only 30 minutes long, so I won’t tell you where in the video the answers are, but the questions will be in chronological order. This means that you’ll need to watch the whole video. However, the first five minutes of the presentation are a discussion of the features that the vehicles have, and the questions will be based on the portion of the video after that.

This one is more technical as Andrej is addressing an audience of mostly PhDs. As a comparison, I do a similar assignment for IS 310, but in that video Andrej is addressing a group of investors and the media, so the talk isn’t nearly as technical. Several of the following questions involves interpreting what he’s saying, filtering the jargon when necessary, and translating his statements in a way that a non-technical person could understand. This is also an exercise to train you to interpret highly technical talks, even on subjects that you’re not familiar with (neural network-based image recognition isn’t my area of expertise), and gather some insight from it.

So, the question: When Andrej talks about a LIDAR sensor pre-generating an HD map, what does he mean? How is Tesla’s approach different?
4. Andrej uses the “stop sign” as an example of corner cases that a self-driving car has to account for. Summarize this problem.
5. How does Tesla train the car to detect these corner cases (like the occluded stop sign)? I’m looking for a few sentences explaining the process, including building the training set.
6. Tesla has a very small team (a few dozen) working on this self-driving technology. Explain how this is possible. As a hint, this does relate to our discussion in Chapter 4 about the outputs of machine learning models. I’m sure I’ll bring it up in class.
7. What does Andrej mean when he refers to Software 1.0 and Software 2.0 code? Hint: This relates to the answer to the previous question, even if this is mentioned a few minutes later in the video.
8. Andrej talks about how (currently) many of the decisions made by the car are done by “explicit predictors” and how they eventually want the car to perform self-supervised learning. Describe what this is in your own words, as if you were trying to explain this to someone non-technical.

Last Completed Projects