15 Comments

  • Dave Ackley

    This is a fifteen minute overview of the essay "Beyond Efficiency", which is appearing in the October issue of Communications of the ACM.  It's a little emphatic!  What do you think?

  • Andy Petsch

    Thanks for this video.
    I love to hear and think about new ideas especially if they break up this way of thinking which takes everything for granted and question the way things work right now.
    You're a really pleasant speaker.

    Please keep doing those videos.

  • Manuel D

    I love watching your videos, you're really smart. By the way I'd like to know if you have a case to share where Quicksort or Mergesort would fails.

  • Stephen Paul King

    Swaps reduce spatial position error… Interesting! This reminds me of an computational analogue of Erik Verlinde's theory of gravity.

  • Stephen Paul King

    Could we think of the Alien Card Comparison Device to a test of the relative entropy of pairs of Alien Cards?

  • gsemman

    Great video. I was in Prof Ackley's programming class back in 2000 when I was a CS undergrad, I enjoyed his lectures a lot. Prof Ackley, do you remember the blocks program homework? We had to move blocks around in a grid until they bumped into something. My friends still tease me to this day how I was banging my head trying to figure out that homework. It's been 13 years now for me as a network driver developer, thanks much to your teaching and much grace.

  • Molecule Magician

    Thanks Dave.  I think I've always lived with or worked around the errors and limitations of other's contributions to a finished system.  I've never thought of system development so much in terms of dynamic and changing interactions of subsystems which may change over time and are created in companies with office politics and social and political tyranny, corruption, misconceptions and false paradigms.  My eyes are more open now.  I think I got it.

  • cyroxis

    Great food for thought. I took a couple of your classes back in 2006 and they were very foundational for both my studies and my career in software. Thank you for educating me both then and now.

  • Niriel

    Have you got any feeling as to the kind of mathematics that would describe the robust systems that you are presenting? Maybe probabilistic logic of some sort, or even linear logic if you assume that performing a computation uses power that may not be available for another computation later. I have no idea.

  • Qwertie

    Err…. a couple of problems here. First, there's still an assumption that the rest of the algorithm runs with 100% reliability – the loops and the swaps – bubble sort is still not reliable if these fail. Second, if you bubble-sort 8192 elements, it'll involve roughly 67 million comparisons versus about 100,000 for merge sort – roughly 670 times as many comparisons, which is rather worse than the efficiency lost from having 3 coffee shops rather than just one! A successful pitch for robust computing must tell us (1) how to actually achieve robustness and (2) how to do it without an overwhelming loss of performance. This is even more true given that our non-robust software has already achieved Google scale and internet scale, i.e. adding robustness as an afterthought seems mostly to work okay.

Leave a Reply

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