ChatGPT proposes and creates a game in love2d ( artificial intelligence scares me )

[knorke]

It is impressive that any somewhat useable code is generated at all.
However I think ChatGPT is just a big remix-system of existing content.
Also it made several mistakes and you basically debugged it together. That is fascinating in its own way. I would like to see an example where the user is less familiar with the task and can not give as good guidance to the AI.

[darkfrei]

However I think ChatGPT is just a big remix-system of existing content.

The whole our life is just remix-system what we saw, what we heard, what we read or what we have spoken for with others.

You cannot make the code without reading it first. You cannot make pictures without seeing of other pictures or real objects of paintings.

Yes, you can make something random, but for explaining of that random you are need other experience (feelings what you saw, heard, read etc.).

[glitchapp]

Amazing! here you have a shader fully written by chatGPT

I tweaked it and made a second version just in case you are interested: https://codeberg.org/glitchapp/fish-fil ... luenZx.lua

[ivan]

What is the license of the code produced by ChatGPT?

ChatGPT is just a big remix-system of existing content.

Yes, and I suspect a lot of it was copyrighted content.

Also, consider how the data collected during conversations is used.
It is quite likely that ChatGPT can be used to profile people online.

[darkfrei]

You can never (!) to get the same code twice, it uses concept of code in any language, not the code itself.

[ivan]

I think there is a broader and more general question of copyright and privacy.
My concern is with regards to the data that is stored and processed by ChatGPT and how we can determine if it was obtained legally.
ChatGPT does NOT cite its sources of information, and should therefore be treated with utmost skepticism.

[GVovkiv]

I think there is a broader and more general question of copyright and privacy.
My concern is with regards to the data that is stored and processed by ChatGPT and how we can determine if it was obtained legally.
ChatGPT does NOT cite its sources of information, and should therefore be treated with utmost skepticism.

Does same should apply to "meat bugs" (humans, lol)? I mean, some people might answer with code examples that they took somewhere time ago, remembered and then post as answer, without citing who is original author. (Maybe even because they doesn't remember where they took it from)
(I'm asking just for sake of discussion)

[ivan]

"A posteriori" statements need to be backed up with sources so we can judge their accuracy and truthfulness.
On the other hand, "a priori" statements are deductive and can be proven using logic or math.
We should not assume that ChatGPT's replies are true.
Moreover, how do we know that the source data was obtained legally?

[GVovkiv]

We should not assume that ChatGPT's replies are true.

But should we assume, that if someone here (at least, on this love forums) reply with code example, that they are posting reply with legally obtained data?
Like, nobody not really question legality of posted code of "meat bag" user, but for chatGPT piece of code it automatically considered not legal...

And, also, can you explain (since you seems quite experienced with it?). If I, user A, give to user B piece of code from person C, that doesn't allow reusing this code, does user C can take legal actions against user B, if they found out about that?

[ivan]

It depends on where users A,B and C are located - each country has its own copyright laws.
Generally speaking, you cannot just copy/paste code that you found on the internet unless it includes an appropriate license.
ChatGPT is not a legal entity and as far as I know you cannot sue an AI.
All I am saying is that we should be especially critical towards its developers.