Hacker News new | past | comments | ask | show | jobs | submit login
ChatGPT prediction:People will type worse on purpose[And why it's a good thing ]
2 points by bulbosaur123 on Dec 9, 2022 | hide | past | favorite
Gonna make some bold, dividing claims from the get-go...

I think we can pretty much conclude that Turing test has been passed. Why?

This why:

https://www.reveddit.com/v/WitchesVsPatriarchy/comments/zfr33a/just_broke_up_with_my_bf_and_it_feels_so_fing/

655 upvotes

and

https://www.reveddit.com/v/AmItheAsshole/comments/zeuymn/comment/iz9hzyj/

210 upvotes

Now those posts have been deleted because someone ticked off the mods that the posts were produced by ChatGPT trolls from /g/. Example prompt used:

https://i.imgur.com/uw7yVsm.png

Sure, we can move goalposts in terms of passing Turing test "No, n-not like that! It doesn't count!", but let's be real here- the threads above fooled hundreds of people enough to upvote, engage, comment and the only reason they were taken down was because OP outed himself.

This has caused increasing paranoia among forum users. I've even noticed that I myself can no longer tell if the ycombinator comment, reddit or 4chan post are real or generated and I found myself reading the posts which look more amateurish, written with errors or in a speedy manner (like the one I'm currently producing RIGHT NOW)

So finally, my predictions:

-> General content fatigue from ever-increasing textual content

-> Increasing paranoia about legitimacy of commentators (are these even real people?)

-> Commentators adopting randomized writing styles (with errors on purpose and just a weird style) - to differentiate from "sterile" AI outputs.

-> AI trolls adopting these changes (injecting some weird randomization in generated content, mistake on purpose)

-> Platforms introducing steganography filters for users to check if user's writing suddenly changes or goes out of bounds (similar to AI checking chess games to see if you make moves within your range of Elo)

-> Gradual transfer to video platforms with "blue checkmarks" where it's harder to cheat (hello TikTok, YT Reels). People will trust real faces and influencers more.

-> New sites like video, audio boards replacing reddit and old-school forums (harder to fake video and audio). Think clubhouse+reddit with histories.

-> Then people will realize they can still use AI to generate content for their video, audio text but AT LEAST they will have some human input and it won't be just bots leading to dead internet theory.

-> Eventually it will be just video, because probably within few years voice generation will be immaculate

Thoughts?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: