AI to help us...

[coilbookmedia]

Hello,

If I make a helper using a mesh object and name it DUMMY_something, your system should automatically make it non-rendarable because it detected the word DUMMY.

Or if I go a long way to get to where I want to be many times  (Menu, for example), but there is a shortcut, your system should detect this pattern  and tell me that there is a short cut for it. (Auto preview a shortcut name  to the side. Eventually we will remember it. Let's I always go FILE>SAVE   auto preview after 10 times a message: "You can also use CTRL+S to save" )

Or if I model/animate it the wrong way, like going through extra unnecessary things, your system should learn my habits and advise a better way to do it.

Or auto-naming based on a  model shape, auto search PC for missing textures because a new hard-drive letter was change.  You still have an issue with it. 

You can add AI  weights how aggressive AI is with suggesting, auto changing settings for a user, correcting user's habits.   

Simulations settings should be also auto previewed  in real time VISUALLY using AI so i don't have to change one number and restimulate and waste my time,

etc. 

[RobH2]

These are all good ideas but since this is a "user to user" forum, the development team doesn't spend much time in here and it's not a direct line to them. There is an "Ideas" forum section. This would be a great place to put this post. All it will do here is get read and nothing will happen as there is nothing to solve. 

If an idea gets a lot of votes, sometimes it gets implemented. Over 70 of the ideas from there have made their way into Max over the years. Maybe yours. I'm sure the Dev team is closely watching and planning for AI functionality so the time is right. Just repost this over there and it will be more useful. 

[coilbookmedia]

Thanks! 

[dmitriy.shpilevoy]

Not sure I want a Clippy in max to mock my ancient modeling techniques.

[RobH2]

Problem is, it doesn't matter whether we want it or not, it's comming. The trick will be to find a way to embrace it and ways to use it to our advantage. I've always lived by the adage that, "the only thing in life we can really count on is that it will change." Nothing in the world stays the same even as badly as we wish some things would. AI is  going to be all over us like a locust plague in the next few years. We'd be smart to jump in and figure it out as we have no choice in the matter...   🙂

[dmitriy.shpilevoy]

My problem with original suggestion is that it sounds like constant monitoring of user's activity and since AIs are quite sluggish atm it would introduce delay/lag and strain GPU. Or even worse - info would be sent to autodesk to compute, which raises security concern.

Imho AI is best suited for doer-tasks on demand: unwrap, UV packing, skinning, retopology. Maybe txt-to-3D and img-to-3D when it gets better. Would be interesting to see autodesk's own model trained specifically for MXS, because it's so niche that everything on the market produces hallucinatory gibberish instead of useful code right now.

[RobH2]

I've resigned myself to the notion that if I'm online, I'm being watched, so I have to just be OK with that for now. I don't have time to fight that battle. 

I too feel like AI is not very useful because like you say, the results are gibberish. I'll post a few tests I did. And here was my workflow and thinking. This is long, so for the faint of heart, you may not read it to the end. Most people don't read more than 3 sentences these days anyway...lol...   But, here goes....

-------------Begin Paste--------------

Instead of being terrified by AI in the 3D world and worrying how much work it might take away I’ve decided to embrace it and see what it can do. So the past couple days I dove in and started working with it via the latest 3D tools. These aren’t done in Photoshop’s AI tools, but instead directly in 3D with TyDiffusion. 

To do it in 3D you need a model. Not knowing what to use I figured I’d use something that wasn’t very interesting on its own. So, I used the dental arch model that we used for a recent dental client project. I probably should have used a sailboat since everything I have on Facebook is sailing related (maybe I need to branch out…lol…). I figured that if I could get something interesting with that, then I could easily get something more interesting with a more interesting 3d scene.

It kind of blew my mind. The way it works, you put key words and phrases in a text box and the AI algorithms use those words to decide what to add to the base model. After the first few, I had an idea, what if I use famous quotes from literature and see what it might create? Then I took it a step farther and used famous song lyrics.

The results are staggering. Some are not very close to the text, but others are amazing interpretations. Imagine what can be done if you actually craft the key words more carefully. I’ll post new ones now and then as I play. But I say, “bring it on AI”…lol…

I posted the actual starting model. It’s just a boring arch of teeth with some gauze on it. I did nothing to the artwork, I just let AI interpret things and I didn’t put the Author or Artist’s name. I didn’t want to help AI figure it out. For the Darth Vader one, the saying is so famous that AI knew it was Star Wars related and created artwork reflecting that. Again, Mind Blowing….

Once we can control it more and it's not as hallucinatory, we might have a useful tool. But for now it's more fun than helpful from a professional work aspect. 

------------------End Paste---------------------------

Attached are the relevant images. 

[dmitriy.shpilevoy]

You forced it to make little blankets for the teeth 😃

I'm surprised it didn't try to add a skull around it, to me that would be default interpretation.

To clarify, it's not that I'm afraid of being watched, but some companies don't like online services. For example, stable diffusion would be approved since it can run locally, but midjourney is a no-no. So if max would add a tool that requires online and uploads meshes to autodesk server, then such a tool would be forbidden for many artists.

At the moment I'm researching practical applications of AI tools, so poking around with stable diffusion, stable projector and a bunch of model generators like meshy, tripo, rodin.

Ran "job stealing" experiment - tried to go through full creation process using AI: generate concept, use it as a reference to generate model and retexture if needed.

So far it looks like my job is safe for a while. It is fascinating and discouraging at the same time - I can see potential in some of it, but most is not ready to be used in production. Though some can give a nice base mesh/texture to build on top.

[RobH2]

I agree with everything you said. Whether we are happy about it or not, we are forced to deal with it and figure out how to leverage it to make money and not get left in the dust. I've reinvented myself 3 times already. Stock Photography killed the photo studio business, so that career died. Turbosquid and CGTrader, both of whom I sell models on, killed my modeling clients so I had to adjust and make models for sale. Now I have to pivot again. 

I guess we learn and grow by having to adapt. Well, isn't that how evolution works anyway...?  🙂  

FYI, I didn't force anyhting for the little blankets. That was a piece of gauze on the teeth for a dental client. I literally just put in the literary quote or the song lyric and that's exactly what I got. I didn't intervene or change the text input. What I didn't do is leave the artist name or authors name to make it easier for the AI. But the AI still knew that "Luke, I am your father" is Starwars and the result reflected that. Kind of cool, scary, amazing and mind blowing at the same time. Like a good roller coaster I guess...

[sebastian___]

What about using AI for max script help ? I noticed people are saying it is useless for that, but in my case it was very helpful. And if you know a bit of max script it's even better since you can spot the mistakes, and it becomes sort of a collaboration between you and the AI (or LLM).

Maybe it's useless for a max script expert, or useless for a complete beginner, since yes, it will make some mistakes.

[RobH2]

@sebastian___  I think that sounds like a good idea. I suspect that some of that can be done already without learning algorithims but on the surface it sounds like a great idea to non-coders like myself. A programmer may have a bunch of opinions on that and I'd like to hear them.