The Nineteenth Byte

2:30 AM

Just gonna put this here

> I'm not interested in participating in the strike, sorry

> I for the most part agree with SE's new policy on likely GPT-derived answers

> The harm a false positive would do could be very high, and they have demonstrated evidence of false positives being possible

> And while I dislike plagiarism, there needs to be solid proof of plagiarism, which I don't believe GPT-detection tools are capable of with enough certainty

> The vast majority of AI posts that get removed are most likely bad enough to be deleted on their own merit anyway.

I think people are kinda overreacting to the new policy on GPT answers, which is understandable given the nonsensical blog post on AI that preceded it

There are three issues with AI answers:
- Plagiarism
- Content quality
- Site-specific bans

Content quality should be handled identically for any answer regardless of its author, machine or not. If it's valid and useful, it's valid and useful.

Site-specific bans are IMO reasonable, since it reduces moderation load, but those are only relevant where neither of the other two factors are involved (it's both properly attributed and of high quality, since if one of those wasn't true you'd either not know for certain it's AI-generated, or it could just be removed on the grounds of quality)

Since an un-attributed AI answer is already against site rules anyway

Unrelated String

yeah the only real argument for blanket ai bans is it makes content quality harder to determine
since a bad human answer tends to have other obvious issues as well

but attribution kinda covers that anyways

Rydwolf Programs

This really kinda reduces to "what should the burden of proof be for plagiarism from GPT"

And I think we should have a discussion about what that is, since the risk of false positives is high, and a "innocent until proven guilty" approach seems very sensible to me as a default

The lack of willingness to have a community discussion about what that burden of proof should be is of course annoying, but hardly off brand for SE, and they at least defaulted to the safer of the two extremes

Rydwolf Programs

2:52 AM

@Ginger ~~I would personally rather the ICBMs be expired~~

(altho really SLBMs are the ones to be worried about)

(and hypersonics but it's not looking like they'll be an issue any time soon)

Rydwolf Programs

3:17 AM

→ 2 messages moved to The Sand Trap

Your wish is my commish

mousetail

4:55 AM

@RydwolfPrograms The issue is moderators already combine multiple tools and get second opinions before suspending people

DLosc

@UnrelatedString I definitely didn't once lock myself out of my computer until realizing I'd switched keyboard layouts by accidentally hitting Ctrl-Shift before typing my password... :P

Adám

7:15 AM

@AviFS I'm not optimistic:

PlaceReporter99

9:16 AM

I have only thought of 22 builtins for my math-based programming language

well, 22 tokens actually

abxyfgΣ∏$¢ť⅟êłỉį+-×÷ř

input a and b
define f and g
using x and y
Σum it all up
and ∏ultiply it together
the $ine, ¢osine and ťangent
⅟eciprocate

łogarithms
ỉist length
įtem in list
the arithmetic operations +-×÷
take a řoot

êxponentiate

Adám

9:33 AM

@PlaceReporter99 Does that mean you are asking what's missing? Do you want feedback on your glyph choices? Something else?

Also, doesn't this belong in the GC?

Hello, World!

10:56 AM

0

A: "Hello, World!"

Commodore C64 BASIC (Non-competing/just for fun), 36 Tokenised BASIC bytes, 34 PETSCII characters 0LIST:REM"{DEL}{DEL}{DEL}{DEL}{DEL}{DEL}{DEL}{DEL}{DEL}{DEL}{DEL}{DEL}HELLO, WORLD! This uses a trick to hide a BASIC line, by starting a REM statement with a quotation mark, you may use the delete ...

Transcript for