Slashdot Mirror

← Back to Stories (view on slashdot.org)

MIT Uses Machine Learning Algorithm To Make TCP Twice As Fast

Posted by timothy on from the do-it-again-for-four-times-faster dept.

An anonymous reader writes "MIT is claiming they can make the Internet faster if we let computers redesign TCP/IP instead of coding it by hand. They used machine learning to design a version of TCP that's twice the speed and causes half the delay, even with modern bufferbloated networks. They also claim it's more 'fair.' The researchers have put up a lengthy FAQ and source code where they admit they don't know why the system works, only that it goes faster than normal TCP."

4 of 250 comments (clear)

  1. All Jokes Aside... Still No. by Jane+Q.+Public · · Score: 5, Insightful

    Allow a computer to design a faster TCP? Sure!

    Let them actually implement it without knowing how it works? Oh, Hell no!

    I'm not talking "Skynet" or anything here... but if it breaks, who's going to fix it?

    1. Re:All Jokes Aside... Still No. by techhead79 · · Score: 5, Insightful

      we don't have to know WHY it works better, just that it does and how to build a working copy

      But the fact that it does work better means we're either missing a part of the picture that is obviously important or the AI version is leveraging quirks with the system that no current model we have represents. I'm shocked to read that anyone would be comfortable just ignoring the why of something just so we can progress beyond our understanding. If we don't understand the why then we're missing something very important that could lead to breakthroughs in many other areas. Do not let go of the curiosity that got us here to begin with.

    2. Re:All Jokes Aside... Still No. by blankinthefill · · Score: 5, Insightful

      I don't think they just drop the questions and run with it. I'm pretty sure that, when we don't understand how things that are useful work, we just implement them... and study them at the same time. I guarantee you that SOMEONE, at least, is studying why an AI antenna works better than our man-designed ones, and they're doing it for the very reasons that you mention. But I think the point the GP was trying to get at is that we've never let out ability to not understand things hinder our adoption of those very things in the past, and as long as we have good evidence that this thing performs correctly, and we can replicate it, then why wouldn't we use it at the same time we study it?

  2. Come on now by dbIII · · Score: 5, Insightful

    As complex systems goes there are far worse. Go ask an engineer or a scientist.