I tried asking chatGPT for gegl filter for GIMP and it started giving me this xml

does this mean anything to you. I just know it's xml and it gave these instructions:
Explanation of the filter:
glow-color: The color of the glow inside the object.
radius: The intensity of the glow, achieved by blurring the alpha mask.
blend-mode: The mode used to blend the inner glow onto the original image.
opacity: The opacity of the glow effect.
Steps to install and use the filter in GIMP:
Save the above XML code as gegl-inner-glow.xml in the ~/.local/share/gegl-0.4/operations/ directory (on Linux). For Windows, place it in C:\Users\[YourUsername]\AppData\Local\gegl-0.4\operations\.

Restart GIMP.

You should now see "Inner Glow" in the GEGL operation stack or through the search functionality.

Would you like to modify or extend this filter? Let me know what tweaks you want!

It's much more than a language version difference. The API has significantly changed.

Yeah it's going to be an adventure for sure. but i will give it my best.

BTW it's great to see you again ofnuts.

I will look toward ChatGPT for help with making GIMP 3 plugins

Good luck with that. ChatGPT won't find how to write these by itself. It will know how to do so only once there are enough examples of code in the wild. I don't expect it to "know" the Gimp2.x/Gimp3 differences until 2030.

I thought it since it has reasoning it can read instructions from documentation and write them I don't for sure though

trandoductin: It doesn't have "reasoning", that's just a marketing term.

I don't know about 2030, but currently its training set contains many examples for GIMP 2.10 plug-ins and very few from GIMP 3.0 (especially since its API is not finalized). Based on the frequency of examples in its training data, it's more likely to produce 2.10 syntax than 3.0. That will change over time as more people make and document 3.0 plug-ins (so it can train on them), but that's the current state of things.

Are you saying they're faking reasoning? hmmm.

Why bother living when AI can do it for us?

I think it's the same as why bother if there's someone better than us. The answer is up to us individually I think and when we're stuck we ask an expert with AI the AI just happens to be the expert in a lot of cases.

Sure - not to say that AI tools aren't impressive, but there's plenty of misinformation too. For instance, the CEO of Google claimed a while back that Bard learned Bengali without actually training on it. It turned out that the model actually had been trained on the language.

Regarding "reasoning", my understanding (at a high level) is that internally it runs your prompt multiple times, weighs the various responses, then picks the "best" solution according to its weights. That can definitely produce better responses, but it hasn't evolved or anything.

hahah runs multiple times and picks the best one....that's funny sounds like the infinite monkeys on keyboards story theory.

Doesn't think AI will EVER be better than people. Even though it may look like it sometimes.
Yes, just like a monkey learning what it has been taught but will never have the imagination or innovation of a human.

This is just very intensive statistics. These things are called "Large Language Models". They don't know that Gimp is an image editor. They just noted that 'image editor' appears a lot more than other things in the terabytes of text that they have used for the training.

Two interesting videos (some math inside, but even if you don't understand that, you'll get the gist):

I'll tell you what. I used to program and run robotic welders.One was an RJ 2 and the other was older, an RG.
I could have programmed them to smack a coworker in the head but i did not and never would have.

They only do what the programmer tells them to do. They cannot learn by themselves UNLESS someone programs them to.That's AI (artificial intelligence).

You know, like a drone that goes out and kills innocent people.

They only do what the programmer tells them to do. They cannot learn by themselves UNLESS someone programs them to.That's AI (artificial intelligence).

How can it learn by itself even if programmed to? How could it be programmed to have human imagination? Not all human imagination is the same or even logical which it would need to be?

We don't know. But if you consider that intelligence is recognizing patterns and reapplying them, it may just be a matter of the "depth" of patterns. For instance, image generators when from the first generation "any number of sausage shapes side by side" to the second "5 shapes, with 3 joints each". We humans recognize the hand pattern in animal paws and claws, and even in dolphin flippers and fishes fins, but also in robots and claw machines.

The issue would be the recognition of what looks good and the desire to improve results. These things would have to come from a human being even if we manage to fully understand and program intelligence and imagination. AI is, and always will be soulless.

That's what is scary Erisian too, which mean's it would probably be merciless.

I have had experience with ChatGPT where sometimes it works flawlessly, other times, it would write code like as if functions and parameters exist when they're all hallucinated by it.

hallucinations in code either breaks it or might cause major problems later down if not tested properly.
Worse case I can imagine is if the physical robots start hallucinating and do stuff that it think is serving people when they're just hallucinations that would be really scary.