Hello! I have a question:
Lets say i want to make a project using ai for some of the stuff. I will take a maze for example. Could i use ai to make the maze path itself, and do everything else manually, such a decorations and making the maze and state I used AI for the maze? In shorter words, i want to use ai to make the maze Path/Layout , not the 3d model of the maze, and do everything else by hand. Is this allowed, and what are the rules for ai? im new, and dont know really know the rules. I looked it up and saw that you have to state the use of it, and where you used it, but i didnt see anything else. Thx!
I don’t think this would be an issue but, as you said, it’s best practice to be open about your AI usage in the description of your project.
1 Like
first off, nice crowdfunding project! that relic rush looks very nice! i will definitely print that, my family loves games. Cant believe you raised so much money so quick! Also, im glad that using AI wont be an issue. thanks! Btw, do know what, if any, rules are about ai? I know you have to state the use and where but i dont know anything beyond that. Thx!
1 Like
It sounds like you are talking about brainstorming the idea with AI. You would still pick the path that you like the best and then make that path in some CAD software. To me disclosing every influence would be hard. I’m not sure I would disclosed that I had AI help me with the path creation unless you absolutely wanted to.
im not using it for brainstorming, i already have ideas, im just using it to help me make something that i turn into a 3d model
I’m pretty anti-ai but having it aid you in that way is probably fine since you’re more just using it like a search engine (or at least that’s how it sounds)
As long as you’re not one of those hueforge weirdos uploading bloated 3mf files of dinosaurs with two navels and five limbs I don’t really think anyone would get on your ass too hard about it.
1 Like
Like an image or the actual mesh for that path?
i would do it for the image
Okay in my mind thats brainstorming. if you are taking the image and modeling it yourself thats way different than creating a mesh and calling it yours.
2 Likes
…and then it gets tricky, what if you got your 3D model done. let’s say a lemon slice. Right, a sphere, a couple of loop cuts, etc. and done in a couple of minutes. Now to apply the texture in, let’s say Blender, you can use a depth map. Find a 2D image of a lemon slice or even draw it yourself, then use an AI tool to create said depth map. put that on your model from both sides and create your texture.
Now, is this okay as well? It was basically an 80/20 because you really need to know what and where to click… the fact that you used AI makes a difference now? I’m sure you can use Photoshop to create a depth map with a simple click of a button… is that AI as well or is Photoshop just a program?
The things is, it’s a slippery slope… is it just AI when you take a “give 2d image, get 3d model” approach or is there a certain percentage? Where should we draw the line? Without the texture the lemon slice would require a lot of imagination. It’d be like clicking on the “torus” primitive in Bambu Studio and calling it a donut or bagel.
EDIT: found something very similar here, just pretend the orange is a lemon.
1 Like
If you 3d scanned that lemon and uploaded that model what would we call that?
In my mind AI is the 2d image to 3d model and post online. No real intervention from the end user.
I would even argue if you would take that 3d model you got and remixed it then I wouldn’t call it AI. It would have to be a true remix but I’d be fine with that. Basically the equivalent of taking multiple primitive objects and reshaping them into something recognizable.
I think we’re entering Ship of Theseus territory here. For those reading not familiar with the thought experiment and not willing to google it; in a nutshell… the original thought experiment asks: if a ship has all its planks replaced over time, is it still the same ship? If those original planks are then used to build a second ship, which one is the “real” Ship of Theseus?
You mean with a 3D scanner or using photogrammetry with your phone? I’d call that a 3D scan. What AI would be involved in this case? I mean unless it’s open source you can’t really check, and that’s only true if you can read code. Technically, your fridge is an artificial intelligence agent… a very simple one, but based on its environment, it “knows” to turn off the lights. I’m not joking by the way, you can refer to Artificial Intelligence: A Modern Approach (Textbook by Peter Norvig and Stuart J. Russell) where they describe different levels of AI agents. When people say things like…
…they’re probably referring to “using” ChatGPT and other LLMs and not about being against genetic algorithms or simulated annealing and alpha-beta-pruning used in code that you can’t see anyways because as a user you’re working with the front-end of whatever program or app. Either way, if you’ve taken grad level classes in AI you can quickly get a migraine arguing the “use of AI” when you slowly realize all people involved in the discussion have all kinds of different understanding/definitions of what AI is.
I’m saying this in general and from experience, not referring to anything said in this thread. But next time somebody asks you if you can explain “artificial” intelligence to them. Ask them to explain “natural” intelligence to you first to ensure you’re talking about the same thing. You’ll quickly find that a lot of them get “intelligent” confused with smart. This is even easier if they have little children and you ask them if they have “natural” intelligence. Then ask how much their 5 year old knows about differential equations or calculus… guess not so smart/intelligent. Let’s not forget that John McCarthy made up the term “artificial intelligence” in 1956 and people would probably feel less intimidated, intrigued, annoyed, or whatever it is they feel had he called it “chocolate cookie”; but I digress…
And of course AI (the actual kind not the media hyped use of the term) gets incorporated into photogrammatry these days… I’ll leave the rest to Google’s AI to explain… just because it’s funnier this way. 
While traditional photogrammetry is a geometric process, AI is increasingly integrated into photogrammetry workflows, enabling automation and intelligent feature recognition. AI, especially through deep learning, enhances processes like feature detection, image masking, and bundle adjustment,making photogrammetry faster and more accurate. Therefore, photogrammetry itself is not AI, but AI is becoming a powerful tool within photogrammetry.
Here’s a more detailed breakdown:
- Traditional Photogrammetry:
Relies on geometric principles and comparing pixel data to reconstruct 3D models from multiple images.
- AI in Photogrammetry:
AI, particularly machine learning algorithms, is used to automate tasks like:
-
Feature Detection and Matching: AI identifies and matches corresponding points in images with greater speed and accuracy than manual methods.
-
Image Masking and Segmentation: AI can automatically identify and remove unwanted elements from images, like moving objects.
-
Bundle Adjustment Optimization: AI can improve the calculations involved in bundle adjustment, leading to more accurate camera poses and 3D models.
-
Point Cloud Filtering: AI can help remove noise and unwanted points from the generated point cloud.
-
Texture Blending: AI can subtly blend textures and reduce visible seams in the final 3D model.
-
Examples:
- Agisoft Metashape: A popular 3D reconstruction software that utilizes AI-powered features for tasks like classification and depth map generation.
- Artec Studio 19: This software uses AI to automatically analyze images and combine them to create complete 3D models, including from videos.
- NeRF (Neural Radiance Fields): A photogrammetry process that uses AI to reconstruct scenes from images or videos.
In essence, AI enhances photogrammetry by automating tasks, improving accuracy, and enabling more sophisticated 3D modeling capabilities.
1 Like