Rule 7, Original Work, and AI

[SirWumpus]

The obfuscated C program must be an original work that you own.

I think a stronger statement should be made. There is no comment about AI generated code (ugh we've come to this), which in theory the author owns BUT DID NOT CREATE.

The obfuscated C program must be an original work ***created*** by you, the Author.

I do not know the Judges' mind when it comes to AI and AI generated C code, but I feel allowing AI generated code would be a HUGE insult to past winners and those who put their hands to keys, use their imaginations, and think.

[xexyl]

How about .. created by a human instead? Since the rules state that one may use code by another if they have permission and indeed this has happened.

I know of one person who even won 'by accident' as someone had used code of theirs!

And yes it would be an insult to past winners. It's a shame. But it would be better I think to specify by human or otherwise not by AI or LLM or anything like that.

[xexyl]

FYI here is the example I referred to: Oh dear. I accidentally won the IOCCC 2012

[SirWumpus]

The obfuscated C program must be an original work ***created*** by you, the Author and biological human.

[xexyl]

The obfuscated C program must be an original work ***created*** by you, the Author and biological human.

.. or another human as the rules do allow that. Unless the judges want to change this of course. Seems a shame if they do (in rare cases I guess only) but it's not very common anyway so probably doesn't matter.

[lcn2]

FYI here is the example I referred to: Oh dear. I accidentally won the IOCCC 2012

Arron Goethe based their entry “in part” on his code (which was not a copy), and disclosed the other person as a co-author. They also acknowledged that Conway is the original algorithm developer.

BTW: Conway was a good friend of Landon, who tragically died early on in the COVID pandemic. Conway was amused by that winning IOCCC entry as well.

[xexyl]

FYI here is the example I referred to: Oh dear. I accidentally won the IOCCC 2012

Arron Goethe based their entry “in part” on his code (which was not a copy), and disclosed the other person as a co-author. They also acknowledged that Conway is the original algorithm developer.

That's true. I did see something about that.

BTW: Conway was a good friend of Landon, who tragically died early on in the COVID pandemic. Conway was amused by that winning IOCCC entry as well.

That is so sad ... and people not taking it seriously still. I know of someone who has had long Covid for over 500 days and that was some time back. Stay safe everyone!

[lcn2]

The obfuscated C program must be an original work that you own.

I think a stronger statement should be made. There is no comment about AI generated code (ugh we've come to this), which in theory the author owns BUT DID NOT CREATE.

The obfuscated C program must be an original work ***created*** by you, the Author.

I do not know the Judges' mind when it comes to AI and AI generated C code, but I feel allowing AI generated code would be a HUGE insult to past winners and those who put their hands to keys, use their imaginations, and think.

We don’t want to stop anyone from using any tools they like when they’re working on their submissions. So, we don’t think we need to change rule 7 in this case.

The IOCCC has a rich history of remarkable winning entries created by authors who skillfully employed various code-related techniques such as code generators, code analysis tools, machine learning tools, natural language models, code copilot tools, and integrated development environments. While we strongly oppose mandating their use in order to compile or run a submission, individuals are free to continue to create their submissions using them.

Just a heads up, there have been instances where the winning entry code was quite different from the original author’s input. We’re also aware of at least one winning author who can’t use a keyboard.

While we acknowledge the ongoing controversy surrounding machine learning tools and natural language models, we have been considering the development of a guideline that explicitly states that their continued use in the IOCCC has NOT and continues to NOT be prohibited. We have even been contemplating a guideline that encourages the disclosure of the use of such tools in the submission’s remarks.md file.

We understand this might not be what you were hoping for, @SirWumpus, but that’s just how things have been and are.

[xexyl]

The obfuscated C program must be an original work that you own.

I think a stronger statement should be made. There is no comment about AI generated code (ugh we've come to this), which in theory the author owns BUT DID NOT CREATE.

The obfuscated C program must be an original work ***created*** by you, the Author.

I do not know the Judges' mind when it comes to AI and AI generated C code, but I feel allowing AI generated code would be a HUGE insult to past winners and those who put their hands to keys, use their imaginations, and think.

We don’t want to stop anyone from using any tools they like when they’re working on their submissions. So, we don’t think we need to change rule 7 in this case.

The IOCCC has a rich history of remarkable winning entries created by authors who skillfully employed various code-related techniques such as code generators, code analysis tools, machine learning tools, natural language models, code copilot tools, and integrated development environments. While we strongly oppose mandating their use in order to compile or run a submission, individuals are free to continue to create their submissions using them.

Good point! Including one of your favourites all the way from 1984, mullender! I forgot about that.

[SirWumpus]

We understand this might not be what you were hoping for, @SirWumpus, but that’s just how things have been and are.

That's fine. As long as its clearly stated and understood from the start of a contest season.

@lcn2 I think you've understood my feelings on the topic of AI. Personally I have injected a lot of time into my past entries, that I would be crest fallen if an AI entry pushed me out. However, as long as we know at the start its possible, then there can be no surprises down the road, and "them's the breaks".

[xexyl]

We don’t want to stop anyone from using any tools they like when they’re working on their submissions. So, we don’t think we need to change rule 7 in this case.

The IOCCC has a rich history of remarkable winning entries created by authors who skillfully employed various code-related techniques such as code generators, code analysis tools, machine learning tools, natural language models, code copilot tools, and integrated development environments. While we strongly oppose mandating their use in order to compile or run a submission, individuals are free to continue to create their submissions using them.

Will add the information from Landon to guidelines as I'm working on them now anyway. Have to find the right place but will add.

[xexyl]

I have made tremendous progress on the guidelines. Unfortunately something odd is happening with the finding invalid links. It seems every link is invalid. Not sure I have the energy to try and debug this right now.

[xexyl]

It must be some kind of bug because the link works fine on the server. So I'll push the change and maybe there is a bug or something funny going on in my environment. Odd though. Why would it suddenly not work when the commit earlier worked fine?

[xexyl]

Anyway I still have to check that links go through to the right place and reread it all but that can't be done today. This specific issue (AI/LLM) should be resolved however. I haven't touched the mkiocccentry synopsis yet because that code has not been updated yet. Tomorrow I hope to look at what was done here and then look at rules.

[xexyl]

Oh I see. It seems it's a display bug (possibly). The test here shows only one invalid link. I'll fix that and see what it shows. Then I can run here and if it's good then we have a bug to fix in display.