44 comments

[ 135 ms ] story [ 1691 ms ] thread
Great idea, Love things like this!

Competition time - Can anyone come up with a title better than "Rust"?

Edit: I've beaten "Rust" by putting in the title " ". Can anyone beat that?! AI clearly has worked out the less substance the better!

Azure ties the single space, as does Perl and also Bezos.
“News” appears to win all of those, but “ ” is still better.
And newt beats " ", because t > s.
(comment deleted)
It seems two spaces " " wins " ". There's no substance, but more of it!

Edit: looks like it's maybe a tie, instead. B wins in case of a tie.

The best I can find is [space] died “ died” it beats rust, java, and even double space
But " ntws" beats " died".
In the spirit of micro-optimisation, “ NtwS” beats “ ntws”
" Show" is 3% better than "NtwS".
Good find, and again " ShoW" is better than " Show" using the capitalisation trick.

  const randAlphaStr = length =>
    String.fromCharCode(
      ...Array.from({ length }, () => 97 + Math.floor(Math.random() * 26))
    );
  const pitRustAgainstRandomFourCharAlphaStrings = async sampleSize => {
    const rustScore = await getPrediction("rust");
    let rustLosses = 0;
    for (let i = 0; i < sampleSize; i++) {
      if ((await getPrediction(randAlphaStr(4))) > rustScore) {
        rustLosses++;
      }
    }
    console.log(`rust lost ${rustLosses} times out of ${sampleSize}`);
  };
  pitRustAgainstRandomFourCharAlphaStrings(100000);
Result:

  rust lost 55469 times out of 100000
I'd say rust fairs worse than the average random four character string with the letters a-z. That should tell you something about the usefulness of this model.
Can you use your superpowers to find us the best 4 character string?
Title A: Show HN: New grep tool written in Rust

Title B: Show HN: New grep tool written in PHP (7% better)

Hmm.

FWIW, Python fares worse than Rust, C++ beats PHP by 2%.

My guess for PHP scoring higher than Rust is that people love to hate on PHP - it's practically the HN national sport!

NB! I'm not arguing for or against PHP, and don't want to take this thread on THAT particular tangent!

Checks out to me. The first one already exists while the second one is hilarious.
Title A: Show HN: New grep tool written in Python

Title B: Show HN: New grep tool written in Garbage (1% better)

I guess Python is 1% worse than garbage. Lol

TITLE A: Someone should sell this- Montessori toys [4% better]

TITLE B: How you can earn $46k/M selling Montessori toys

Tried it for on of our ProdcutByte products

(comment deleted)
Please don't use this in practice. It's trained on a dataset that it then ends up affecting, so we could end up with some really weird situations.
From their README:

"Hacker-AI uses a brain.js feed-forward neural network machine learning model that has been trained with the below dataset that contains all stories and comments between 2006 and 2017...The current algorithm has been tested with approx. 10.000 posts. It has been able to predict 60% of the cases correctly."

I love it when people come here and shit about old data sets or other bs. Gj op! Haters going to hate and makers going to make.
Gay is 23% better than Straight. Is HN a gay hub?
I want to know how many titles they A/B tested to get this one.
Seems to be affected by heavily by capitalization in a surprising way. "Show HN: AI OpTiMiZeR ToOl" fares 6% better than same written in non-braindamaged way, and fares even worse when compared to all caps.
> The title is transformed it into an array of numbers between 0 and 1 (each character is a number) and this data is fed (along with the number of points) to the neural network.

This sounds strange. But after looking at the authors bio, it isn't strange any more. This has less to do with AI and more with cheap growth hacking.

Yeah, this is a classic case GIGO.

> Title A: kdjfs;dfisd;diofuosd

> Title B: sdlkn nmkd kljdsa;'[ (10% better)

> The winner is title B: sdlkn nmkd kljdsa;'[

I had to look up GIGO; "Garbage in, garbage out", https://en.wikipedia.org/wiki/Garbage_in,_garbage_out .
Converting text to numbers via individual characters destroys any semantic meaning to the input. Which is why no advanced NLP tools will do such a thing. Basically, what they describe takes whatever meaning is in the title and throws it in a shredder.
It's also strange because what they've trained is essentially a recommender system. However, at inference time you're not trying to rank 2 different articles, but the same article with 2 different titles, which I guess you could do, but I'm not sure how well this correlates to your actual ranking on HN.

I guess something more interesting would be to take into account the current front page and see how your article would rank against those, or something along these lines.

In any case, neat idea.

The predict button doesn't work with the default values.
A: "Show HN: I built a decentralized data set index" [1]

B: rust

=> B is 66% better.

Damn it, I should have used rust! Anyways, great project. Bookmarked!

1: Shameless self-promotion: https://rugpullindex.com

A: "We predict the future using Erlang VM, and quantum AI blockchains."

B: Rust

=> B is 71% better.

Rust is unstoppable, clearly.

Hahahah great meme:

A: We cured corona virus

B: rust

=> rust 64% betteer

I'd just like for HN to have the ability to add a 1-2 line summary with the title - a lot of titles feel like clickbait.
On an iPad (iOS 12), the predict button does not work at all
I don't know about this tool. Everything beats "JavaScript React Hooks to organize your habits", including "PHP".