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

Then I realized it's all about my abiltiy to filter there.

There is a somewhat small but vocal and obvious contingency of HN posters who are just repeatedly arguing in bad faith or making low effort comments.

They are also often found discussing almost exclusively political topics and have very little to offer in terms of technical expertise.

I think my experience would be massively improved by being able to filter them out.

I see filtering as a kind of distributed community vote, and eventually if the majority of the community starts to filter out certain members, the net effect is like a distributed shadow ban. Those people won't be engaged with and will probably leave for somewhere they can get reactions.

I'm guessing this goes against HN culture and won't be accepted, but I thought it was worth discussing, because my ability to filter has left me with a pretty delightful experience on an app otherwise not known for them, so imagine what it could do here on HN.

Comments URL: https://news.ycombinator.com/item?id=34229448

Points: 4

# Comments: 2

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.

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.

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.

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.

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.

> 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).

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.

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.

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.

I wonder to what degree certain parts of diffusion dictate using certain noise functions, and how much this paper truly challenges how we understand them. Cool to see it was researched.

Next idea: it seems like a lot of steps could be skipped by using things like momentum during the inference time. I'm sure OpenAI has already implemented several clever tricks like that in production for DallE.

I'm working on (various, non-diffusion) methods for 2D drawing to 3D output right now.

1. "Either everything is magic or nothing is", and magical statistical tools are currently some of the coolest magic we have harnessed.

2. Making realistic pictures of "Corgi's with sushi" is cool.

3. Papers describing architectures which can make ever more realistic or interesting pictures of corgis and sushi are deserving of academic recognition, even if they can't precisely describe how, just as renderers which can do so would be.

I've found a lot of papers which primarily study the theory of ML coming out of the UK, A yet very few of them end up being of much value to the advancement of the field in terms of applications.

Conversely, the people focused on making real applications ("tools") seem to also be innovating on the science.

Tesla and Edison contributed massively to the progress of the study and application of electricity and both spent nearly no time in academia and instead focused on practical applications in industry. Edison's methods of investigation were said by Tesla (somewhat admiringly) to be entirely empirical. I wonder if the UK's strict academic approach to ML may be holding them back from making bigger contributions to the field. I'm glad some are trying to be strictly scientific about it, but I'm also glad it's not everyone, because it doesn't seem to be what's delivering the "magic" we all want.

And if they're right, well then what's there to be upset about?

Distill.pub for example, despite being written by people who worked at OpenAI and (also funded by) Google Brain, makes nearly no mention of either. They've managed to successfully accomplish their goal (creating resources to help people improve their intuition about machine learning) without resorting to those tactics.