Comments & Discussion

Boston University moderates comments to facilitate an informed, substantive, civil conversation. Abusive, profane, self-promotional, misleading, incoherent or off-topic comments will be rejected. Moderators are staffed during regular business hours (EST) and can only accept comments written in English. Statistics or facts must include a citation or a link to the citation.

There are 8 comments on Self-Taught, Self-Driving Cars?

  1. I’m totally amazed that safe driving techniques can be learned by observing other cars. In Boston, I would think the results would be just the opposite!

    1. Almost anywhere. Here in Texas a lot of people drive in the left hand lane when they should not be. When my dad first visited our family in Texas he mused that to drive fast in Texas you needed to be in the right lane!

      1. Great points! Robotic systems are still in their infancy – the more data and experience here, i.e., of how to act or *not* to act, the better. But yes, a student copying their teacher is only the first step, and is not always advisable! Moreover, in practice the other surrounding vehicles can also be another company’s autonomous vehicles, so our system can “level the playing field” in some sense in addition to improving overall safety.

  2. As a retired flight test engineer and having done some System Safety Analyses, how do you keep the car safe when a system fails? I do not believe in driverless cars. I have seen too many automatic systems fail, and the pilot have to shut them down and take over. I want nothing to do with them!

    1. I agree, as current systems are not yet ready for prime time, gaining trust from people may be a gradual process (trusting too much, too soon can also lead to problems as well). Lessons learned from aviation systems are highly relevant to developments in autonomous vehicles here. In the mean time, due to the fundamental issues you describe, I do think it makes more sense for developers to collaborate, share, and be transparent, as they are creating safe/robust/scalable systems, which was our goal for this work.

  3. They get to their destination safely 92% of the time… so if you commute daily to work, you would have typically an accident every two weeks! They consider that a good result??? I sincerely hope their success rate improves drastically with more training, but the article does not indicate what is planned nor what is required to improve that number by magnitudes.

  4. As a current Undergraduate looking to get into Machine Learning somehow, this article was extremely interesting. About the point , “Estimating the viewpoints and blind spots of other nearby cars to create a bird’s-eye-view.” I thought was the most eye catching. Maybe I am too young but I can’t imagine how the computer can get a birds eye view on it own. What if you were to place small little devices throughout, let’s say New York City, where a signal is caught from a self driving vehicle and then that signal is sent back to the car. As a result, somehow helping the computer painting an image of every cars blindspot.

Post a comment.

Your email address will not be published. Required fields are marked *