This article by Cory Doctorow that someone posted in one of the other AI threads actually talks at great length about how changing copyright laws helps large companies, not artists, writers, etc. I highly recommend the article. It’s very well put together.
3 Likes
Indeed I fully agree and I’m not confident AI won’t strengthen the copyright powers of large corporations
2 Likes
Right idea, very sound legal stance. Theres no point in taking unneccessary risks.
I am of the opinion that most if not all LLM’s alone are nowhere near the level of good reliable compliant software required for important programming.
As a judge of a case involving an LLM and code copyright i would ask for proof that the LLM produces or produced a complete copy of the code and proof that the code isnt some fundamental law of logic /physics or whatever or some way of handling something that is unavoidable.
I would take exactly the same steps if a human had done the same.
They, the developers, obviously havent sorted out a training set good enough for general use.
If i was an AI developer with any kind of customer base i would make a software generating machine that follows simple repeatable steps that a human could also follow …
I.e.
- Write a description of the required software.
- list the requirements
- Locate the objects (from nouns)
- Locate the methods (from verbs)
- Implement and test the code
- Refactor using stylistic and logical optimization rules
- Test again
- Deploy for human testing
A Human could also follow those sort of steps, and could produce similar results. What would happen to the program for the following …
“The Supplier sells to the Online Shop who buys for half price then sells to Customers using the internet Shop”.
the implementation of ‘buys for half price’ is hardly likely to differ all that much from a GPL equivalent.
Another reason to avoid LLM code is LLMs Create More Bugs.
3 Likes
In the end my personal philosophy is:
If you can contribute to the engine by your self, do so, if you can’t, don’t
If you need some tool like an LLM to at all be able to create a PR then you can’t really be relied upon to actually understand and update the PR based on feedback and changes, nor can you take long term responsibility for the changes you made
But if you can make the changes yourself, then there’s no good reason to use an LLM, and a lot of reasons not to
5 Likes
(If anyone is disappointed in AI exclusion)
There is nothing stopping anyone from forking Godot and going to town on AI with that fork.
The results might even be interesting.
2 Likes
A copyright is easy to ignore as it is. Staying on the topic of Godot contributions - who really does expects us to check, and less so prove, that the code we write is unique and does not already exist as a part of a copyrighted codebase somewhere? Anyone would find such a requirement, and the idea of possible repercussions, absurd - unless someone mentions AI that is, then suddenly it flips on its head.
Then still, what logic renders someone an IP thief simply by association, without them intending and acting to steal. Neither law nor human idea of justice work that way. I see this as a purely virtual possibility, made up and pushed by those with an agenda.
What’s disturbing is that they will bring up every remotely relevant reason trying to convince you against the use of AI, without ever naming (or maybe even realising) the personal motives that drive them to do that in the first place. Maybe there is a good faith reason behind this manipulation, can’t make out what it might be though.
If you don’t use an LLM, there is no reason to check any of this because you wrote the code yourself.
You do not have to know about a law or have intended to break a law to be in violation of it and be punished for it. Even if a law requires intent, we aren’t talking about reality - we are talking about what can be legally proven. Avoiding a lawsuit is also important for a non-profit open source project like Godot because if a big company decides to come after them, the legal fees will put the organization under and kill the project. Patent trolling has been around for decades.
This is what’s know as a straw man logical fallacy.
5 Likes
Indeed we have no ability to deal with such a threat, and we have to do our due diligence to prevent it
4 Likes
I noticed a couple of interesting points that were raised during this topic.
- Steam’s policy of disclosing AI use.
- The legal liability of the Godot Foundation.
These are points that might warrant their own forum topics to discuss further and flesh out.
I agree with the below statement.
I know I don’t have the skills to contribute to the engine. As a result, I would never consider using AI to generate code or assist me in coding for the engine, as I wouldn’t fully understand what I was contributing.
If I ever do develop the skills to the point I believe I could contribute to the engine, I would no longer need AI to assist me in doing the work.
For developers that may never go to the GitHub repository and find the section on contributing code to the engine, should there be a more prevalent notice on the main Godot website about their position on AI generated code in the engine? Maybe on the download page?
Even if it they aren’t sure if there is any AI generated code, stating that they don’t know allows the user to be informed.
Almost like the warnings on food labels that say it might contain certain allergens, even if they aren’t ingredients, because food containing those allergens are processed nearby.
3 Likes
I think increasing the discoverability of the contribution instructions would be very good, I feel a lot of contributors don’t see them and it’d help with a lot of other things too like usage and formatting requirements
3 Likes