Alita - Geometric AI

Alita addon for using geometric AI



And how we can use it?
We write Alita and it appear out of nowhere?

1 Like

@persimoon maybe it’s an idea to post your work once it is available :wink:

1 Like

Nothing in the website

1 Like

Now available at food4rhino.

@persimoon what does the AI of the name mean? The plugin does classification but using rhino and custom methods, none based on Artificial Intelligence algorithms, or am I wrong?

1 Like

there is a discussion, contemporary AI is based on some algorithms like Genetics or Neuron System, machine learning. Genetics algorithm is based on Numerical Computation like Newton-Raphson metods or interpolation. all of three methods is coming near to adequate response. and in many times finding adequate response is impossible, so we use percentage. if we focus on some AI application, let me breaks it just to Visual Recognition and Geometric Recognition (primitives, something that humans knows) if you make a section of bottle, you can recognize bottle based on its geometry. there are some advantage for AI based on geometric algorithms vs raster discrete computations:
1- it is very very faster
2- it results adequate solutions (and this is very important for autonomous vehicles and drones)
3- it uses a method based on human brain and eye,human can understand AI network brain and absolutely AI understand human. it guaranteed peace between us and robots.
4- we can defines beautify and sensation based on geometric relations recognition
Floral Symmetry
and many more


That’s hilarious.


There is a discussion about what is AI and what not, because it is a multi-disciplinary intersection and what was smart yesterday may be normal today. The main families or AI are machine learning algorithms (neural networks, support vector machines…), agent system algorithms (robots, boids, genetic algorithms,…) and graph theory algorithms (network analysis, liquid state machines, …). The classification task is solved by neural networks, but you are using explicit programming (the conventional way of write code, rather that let the algorithm learn the solution) using decision trees, which is fine and I’m interested in anyway, I just don’t see how this fits into the AI. Classification is an AI task because it is solved using AI methods, not because it is the classification task.

Machine learning is fast because it needs a previous step (training) that is terribly slow for real problems before be used.

That is not true and contrary to its usefulness. If the neural networks learn too well it is called overfitting and prevents the generalization of what has been learned, which is what is sought. Autonomous cars in particular are an example of the fact that machine learning alone is not enough for error-free solutions.

AI understand human? :face_with_raised_eyebrow:

Geometric AI for me refers to AI methods that have geometry as inputs, or that model directly in latent space. I am going to the automatic learning of product design and aesthetics for the next version of Peacock. And believe me, you won’t be able to classify something as complex as aesthetics or beauty using explicit programming.

All this talk of including AI in the name is not meant to invalidate your work, but it is important to call things by their name to avoid miscommunication. I’ve tried your plugin and found it promising and frustrating to see that it doesn’t use AI techniques (I don’t consider the search tree classification an AI algorithm, even if it’s used a lot).

Anyway, how do you expect to classify the symmetry as in that picture? I’m very interested in that problem.


I appreciate for your comment. I’ve tried Peacock. you described complete and thanks.
but 2 cases about your comment for now:
1 - Peacock is a complex generative algorithm like yellow and more. tell me after final design and after some steps can you prepare it for fabrication?
we need a geometric identifier to do that. I saw many question for this case, we need to classify particles.
2- every complex can be classify based on equal shapes,same shape,like shapes, same topology and more
let us try more and thank you if you’ve check our next beta version

I know this is not for any data, we work on a tiny section of geometry. something little but we start to make in applicable

Thanks Dani

damn, dude is getting ripped over the name :frowning:

1 Like

I see you’re ignoring my point, all right. Since rhino has the algorithm that approximates a line to a set of points using linear regression (I guess) it should be called Rhino AI, to take advantage of the hype for marketing.

You don’t have to recognize the geometry if you’re the one who creates it :ok_hand:

This is your answer to how to classify symmetry? Because the topology of a curve is invariant to rotations (to affine transformations).

1 Like

Name is important, one should not mislead his/hers audience.

btw, I know they put genetic algorithms alongside AI, but I disagree.


true, but people seem to be a little hyper-critical imo. he obviously spent a lot of time making it available for us and nobody is talking about the functionality. i don’t have v6 so i can’t check it out, but it looks really good for pattern makers.

He probably took it down to rename it.

1 Like

The link he sent doesn’t work, this one should:

I was on a technical conference recently and it was all about AI, VR, AR, Industry 4.0, etc. People are trying to sell everything good and bad and since most software is negotiated on management level (i.e. with people who, usually, have no clue about the technology), very often the “little guys” end up struggling when the software doesn’t meet the expectations.

The name, description and pretty pictures MUST match the actual provided technology.


The name, description and pretty pictures

and random movie poster as the software’s icon (might be copyrighted) :smiley:


Your comment definitely is important, because you are great at generative programming (peacock and hyllomachine). give us 3 months to grow it and let us make it complete.
but about Topology, topology is same for Rotation,Moving and Scale in geometry, I know what you mean, it is different in Rhino NURBS and this is a not a good spot for Rhino and must be fix. in Rhino however topology is same but sort in different.

we are focusing on identifying geometric relations, we categorized this for Classic Architecture, Mimetic Architecture, Tiling and more (total 9 categories that have formula for beauty evaluation.

most of our problems is programming, because we are both are BIM manager.
we need help on implementing algorithms on Grasshopper python and Dynamo python.