MathYouF

"Bachelor's degrees have become essential for well-paid jobs in the US."

The lies we continue to allow ourselves to tell as a society.

The real moment of truth will be if any models start to assist massively in research in the hard sciences.

Based on the quality of outputs I get when asking for help with somewhat complex AI research problems, I think it'll likely help accelerate the pace of other research as well, and discovery will be limited by people's speed of running the tests it suggests and feeding it back the results.

The main value of using SDF shapes for 3D modeling workflows is you don't have to worry about topology (the vertex, edge, face graph structure which has to be formed over the surface of all 3D models) which makes a lot of modifiers (like boolean combinations of intersecting objects) vastly less tedious (Womp calls this feature "goop").

Right now Blender work still involves a lot of tedium, mostly related to topology. A lot of upcoming 3D ML applications also work considerably better when using SDF instead of mesh representations. I wouldn't be surprised to see this form of 3D modeling take off to a significant degree because of those two factors.

The cost of materials and transportation and time using the expensive CNC machine will be the major costs of sculpture. Generating the same quality 3D models is at the very furthest 18 months away. And animating and rigging the models and giving them auto-generated RL policies will surely come very quickly next.

> I don't think AI will be able to replace human creativity for discovering new paradigms as fast as it will replace human application of existing paradigms. And by doing the latter really well with AI, we're killing our ability to do the former. We'll end up with a sterile art trajectory.

This may actually end up making the few artists creative enough to create bold new art styles even more valuable, if they can basically not release their art and hide it behind a model.

Though I guess anyone with access to that model's output could then just generate a few samples and train on those, so maybe not.

With your expertise and the fact that they're so early, you should be able to build a competing business with your superior product right?

Then you can keep all the rewards (and find out all they had to do to be in the position to be able to hire you).

Sounds like a win-win, you should be talking to startup accelerators not startups hiring.

> Numeracy is just as valuable as literacy.

I actually think this is likely not true.

If you were to take an 18 year old who can't read, and another one who can't do addition, which do you think is less employable?

There's a lot of jobs one can do without any math. Almost none one can do without any reading.

There's more to it than all this of course, but I think literacy is the clear winner compared to numeracy.

The tone of this betrays a possibly more argumentative than collaborative conversation style than that which I may want to engage with further (as seems common I've noticed amongst anti-connectionists), but I did find one point intersting for discussion.

> Parameters are just the database of the system.

Would any equations parameters be considered just the database then? C in E=MC^2, 2 in a^2+b^2=c^2?

I suppose those numbers are basically a database, but the relationships (connections) they have to the other variables (inputs) represent a demonstrable truth about the universe.

To some degree every parameter in a nn is also representing some truth about the universe. How general and compact that representation is currently is not known (likely less than we'd like of both traits).

If greater parameterization leads to memorization rather than generalisation it's likely a failure in our current architectures and loss formulations rather than an inherent benefit of "fewer parameters" improving generalizaiton. Other animals do not generalize better than humans despite having fewer neurons (or their generalizaitons betray a misunderstanding of the number and depth of subcategories there are for things, like when a dog barks at everything that passes by the window).

The team I work on is building tooling with exactly that in mind, making this a part of an artists workflow rather than any sort of replacement.

I know as much about how to get the best image outputs from text inputs as the person who designed an airport knows the best place to eat in it. The emergent properties of the system are a result of the data put into it, so I can only discuss the system itself, not what it ended up doing with the data in that system.

The models are a product of their datasets, specifically the relationship of the images and prompts via CLIP. CLIP puts both images and text into coordinate space, imagine just a 2D graph. It tries to assure that for any real image and its caption, they will each be each others closest neighbor in that coordinate space.

So if you want a certain image, you have to ask "what caption would be most likely and most uniquely given to the image I'm imagining".

I'm sure this advice is way less helpful than what you find in prompt engineering discord channels and guides I've seen.

Pixel values are discrete (length x width x r256 x g256 x b256) and vertex values are continuous, so that is one major difference.

Secondly, there's vastly more labeled image data in the world than 3D data, so creating a CLMP (contrastive language and mesh pairing) model is harder.

It's very late but I may be able to give a much better answer on more of the nuances of 3D generation tomorrow.

Like anyone deeply in a field I know maybe several thousand people who could probably give a better answer, but I figure I'll give an effort to provide one since I don't see any good ones posted yet.

The moment everyone knew this was going to be big was in 2019 when StyleGAN came out. They used a lot of tricks like aligning face features (like eyes) and had all their pictures of a single domain (the most famous being faces) but none the less, that was the moment everyone in the AI field knew this was going to be big, and so three years ago a lot of big people shifted to this line of research.

The four main innovations since then have been:

1. Transformers

Generalized computation kernels which allow for images to consider non-localised relationships between pixels of an image. Released in 2017, and originally used for language.

2. Pixel Patch Encodings

Different resolution semantic and geometric image information encodings which allow for better representations of relationships between image areas than pixels are able to achieve given the same compute. Allows using Transformers on high resolution images.

3. CLIP

Contrastive Language and Image Pairing. Before, the only way we knew to classify an image was as a "face" or "cat" or "ramen". When the genius idea of labeling images as semantically meaningful vectors rather than one hot encoded classes was revealed, it changed everything in computer vision very quickly, and problems that used to be hard became trivial. Released in 2021

4. Diffusion Models

GANs penalise you for making an image which does not seem to be part of an existing dataset. This encourages one to make the worst quality image that looks like a member of that dataset. Diffusion learns to denoise an image, removing noise is perceptually similar to increasing resolution, people like images that look that way. There may be more people with better intuition about diffusion models may be able to add on why they're superior. I've read all the papers leading up to the latest unCLIP (Dalle2) but it's complicated. Released in 2020, with major improvements to the training process continuously being made since then.

Hope this was helpful. All of the above were only implemented for images in any real way in the last three years. Putting them all together is something many people only just this year did, resulting in DallE, Stable Diffusion, and Imagen.

I'm working on doing this for 3D and later for use cases in AR. 3D generation still hasn't been cracked the same way image has but the above will likely contribute to the solution to that as well. Anyone who's intersted in working on that feel free to message me.