rbm

AI

generative model demo

SHOW HN2016

145points
49comments
#234of 2,264 in 2016
Top 11%of 2016 by points

How HN took it

MixedNo clear lean

17 positive, 19 neutral or asking, 0 mixed, 3 critical, out of 39 comments about the project. Thread score +0.34.

Idea: 12 for, 3 againstDocs: 3 for, 0 againstDesign: 1 for, 0 against

The thread, comment by comment

Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.

positive neutral or mixed critical about another commenter or something else the maker

What HN said

What people liked

  • Generated names are funny and often plausible

    From 3 comments
    • I love how close some of these are to plausible. "HelloWaurd", "ruby-franwork", "Contect-Server", "php-squere", "griend" (friend + grind?). Also: "libdog". A library, for dogs.…
      camtarn on HN
    • Yep, I got tons of normal-looking ones as well - had to dig to find some of these, after I found HelloWaurd and was tickled by it. The thing I find interesting and amusing -…
      camtarn on HN
    • Pretty nifty! Less than a minute's worth of click reload yielded some pretty good ones.
      zeveb on HN
  • Blog post and visualizations are excellent

    From 3 comments
    • The post behind it has the show-stealing line, and it was written by the author. > Practical Applications > None whatsoever. Excellent write-up, and great visualizations of the…
      spdustin on HN
    • I love this, the blog post is a great write-up. Not a trivial algorithm to implement (although a theoretically interesting one), kudos!
      dasboth on HN
    • The blog post is really excellent.
      erubin on HN
  • Hidden unit zoo is a helpful window into the model

    From 1 comment
    • I really like your Hidden Unit Zoo here http://colinmorris.github.io/rbm/zoo/ as a window into what this thing is actually "thinking" about. The "top matches" for a given hidden…
      wxs on HN

Objections

  • Training data included many non-JavaScript repos, lowering quality

    From 1 comment
    • I wish this was restricted to just the 3.9 million repos of the JavaScript build systems and task runners. The remainder of the repos just tainted the training set.
      IgorPartola on HN
  • Not a new idea; Markov models and char-rnn did similar things

    From 1 comment
    • It's not a new idea. Since char-rnn came out, people have been using it to generate music, generate fake linux code, write Shakespeare, make chatbots, etc. And before that people…
      Houshalter on HN
  • Some names are just spelling errors

    From 1 comment
    • Your examples just look like spelling errors ("contact", "framework", "Word", "square".) I got a lot of ones that looked indistinguishable from real repos, at least on first…
      Houshalter on HN

Requests

  • Add options for language, repo category and similar filters

    From 1 comment
    • it will be great if you can add some optional features to the interface, like the programming language(python, ruby, java, etc.), the category of the repo (game, development,…
      criloz2 on HN
  • Generate names from startup domain names with availability checks

    From 1 comment
    • Someone should build something like this, based off startup domain names, cross-checked with availability.
      whatupdave on HN

Questions

  • Are generated names guaranteed unique?

    From 1 comment
    • I was thinking about the exact same app. Nice. Is each generation guaranteed to be unique ?
      spriggan3 on HN

Category, keywords, summaries and labels on this page were made by AI (Thread Reader v2) and not checked by hand. Each point links to the comments it came from.

Add a badge to your READMEShown on HackerShow
[![Shown on HackerShow](https://hackershow.org/badge.svg)](https://hackershow.org/show/12210212)