That’s what this site is. You’ve been talking to a machine all this time. Do you really think a person would have this much time to argue day in and day out over systemd? That that many people care wherever or not Go has generics and which other languages does them best? That rust somehow is the cure for everything from game engines to the common cold? Think about all the times a “person” stated something was Turing Complete and how it made zero sense to you. It’s been machines all this time.

Yeah, but at some point those machines would have had to have original training data, so at some point people probably did have time to engage in techno religious wars. Unless of course, all of this is made up, to keep us occupied to not discover the matrix or alike ..

Wir sind die Roboten.

Obligatory irrelevant typo correction: it's "Roboter"

Point taken and indeed it is obligatory and relevant. That ‘n’ didn’t sit well with me at all.

Who could argue for hours and praise and upvote someone getting Doom running on all their household appliances?

Ok that’s it then it’s time to run doom on my toaster, rendering onto bread… input will be tough though

Have you considered a touch screen toaster?

https://www.cnet.com/reviews/revolution-r180-toaster-review/

That's a really interesting idea. I think it would be fun to see how well it would work in practice.

https://imgur.com/4UZUPec

Rats, top comments will be impossible to beat. I can probably piggy-back off a top comment, though, those comment threads aren't long yet...

> Possibly, the model could be trained on the actual content of the link. But I suspect just using the short title might be better. Larger concentration on the actual material there, websites contain lots of text which is useless.

Most people do not read the article before commenting, you do not have to train the AI to do that.

I love this idea—I think it would be fun to gamify it! Keep the predicted thread secret, and then use it to award points to hn commenters based on originality. If a comment is functionally the same as a predicted comment, it gets -10 points. If the comment is novel, it gets 10 points. Then there's a sliding scale between.

You'd need some checks to prevent people from posting a bunch of Zalgo text (maybe a model to identify "interesting" or "insightful" posts), but it could be a fun overlay on the hn commenting ecosystem.

Can't tell if this is a parody comment or not...

Those comments violate the HN guidelines.

Please read and follow them: https://news.ycombinator.com/newsguidelines.html

Kidding, I too thought this thread was the parody at first. After all, they are real HN comments!

Or maybe this is part of the parody. Maybe if we trained an AI to detect genuine vs. parody?

> Maybe if we trained an AI to detect genuine vs. parody?

Blockchain will solve this better

There was a reddit experiment a few years back, SubredditSimulator, that was a scaled version of what you described. It was obviously more surreal and funny than accurate. There were users trained off of popular subreddits, and they just ran with threads and comments being created purely from the data.

Lots of us have had that idea :) we should band together and make it happen.

Let's go a level further. Instead of banning/shadowbanning people just send it to shadow realm where all they see are bot created comments and bot created answers to what they comment

Gold

See also: heavenbanned

https://twitter.com/nearcyan/status/1532076277947330561

Some video games already did something like that, except instead of AI they put you in match with other offenders, in sort of personal hell, then only let you out once you won some matches

Sounds similar to /r/subredditsimulator.

>this is a fully-automated subreddit that generates random submissions and comments using markov chains (see below for more info), with each bot account creating text based on comments from a different subreddit.

Quite fun to read.

Someone did something similar before, but unfortunately it's now offline. The HN link is still up, though: https://news.ycombinator.com/item?id=19694578

I present algorithm whereby each byte is iterated from 0 to 255, following recursively for next byte for each previously yielded result. From above nonsensical results are removed.

Above algorithm generalizes all human knowledge ever conceived, in any format, as well as any future (duplicated!) ideas that may be conceived.

I donate above algorithm and all its results to public domain.

You're welcome.

> I present algorithm whereby each byte is iterated from 0 to 255

You should've just said a stream of bits, as your system cannot generate anything that requires a non-multiple of 8 bits to be accurately expressed. Zero padding is not the same as absence of bits.

I did this quite a few years ago. If I remember correctly, it was so convincing that people emailed Dang because I set the username at the top to be “pg”

https://news.ycombinator.com/item?id=10248773

Do you have your sim archived anywhere? The original seems to be down.

No, but I’m building it again. This time in rust. There’s some stuff on my GitHub to download data in bulk from the HN API

> but I’m building it again. This time in rust.

Good crab.

> websites contain lots of text which is useless

I'm honestly really curious about this. Could you elaborate?

Headers full of navigation. Sign up for my newsletter spam. Footers. The webpage isn't just "the document we're talking about", but an entire application most of the time.

I'm also surprised someone hasn't made a parody LLM-based HN thread generator yet. Perhaps until recently they were too busy writing crypto startup business plan generators to bilk gullible VCs.

There are GPT-3 generated subreddits

They are pretty spot on and could do this with hackernews