andrewnc

Hey this is great! How long does it take to generate the Wiki when setting up, and what have you found is the best way to keep it up to date?

I think "The Gong Show" was an old tv show about amateur talents. Sometimes good, most of the time terrible and hilariously unaware. Not sure if that was what was intended here.

Yes! It definitely should be

My favorite part here is that it's a noun and a verb

"i-kwan mo yung kwan" -> "just thing the thing"

Blog post is out now https://gretel.ai/blog/introducing-gretel-amplify

They get 43,300 records per second on this example, which seems to the right order of magnitude for you

We have some preliminary work in this direction https://github.com/gretelai/multi-table

I love the idea of "table space" though. It would be fun to traverse this space and output a new database at each step, like a VAE.

We're currently beta testing something that fits this use case directly. The models we have today are really great at capturing the original distribution, but they're not always the fastest. This new stuff will change that, feel free to reach out (maybe on our slack?) and we can see if we can get something working

We've been working hard to make it super simple to get started with useful synthetic data.

If you want to know how you would go and use this for your own problems check out some of our other posts

https://gretel.ai/blog/how-to-safely-work-with-another-compa...

One examples of this is back translation [0]

It works fairly well as model size increases

Also, shameless plug, we're pretty proud of our generated anything's at Gretel[1]. It's tabular, text, and time series for now - but we recently had a blog post that shows how generated data can be useful for downstream ML [2]

[0] https://arxiv.org/abs/2110.05448

[1] https://gretel.ai/

[2] https://gretel.ai/blog/how-to-safely-work-with-another-compa...

I saw a few DALLE coverarts in there ;) very cool!

Thanks for the feedback, I was hoping the course would be something of a redemption.

I self published the book to prove I could do something like that. It got much more traction than I had planned and, in hindsight, I wish I had paid for editing and formatting as a minimum.

As for the quality of the paperback, that was unfortunately out of my control as I used Amazon's print on demand services. Definitely a painful lesson for me.

In any case, I appreciate this comment and others here. I'm definitely working towards much higher substance with increased polish. :)

I've been a long time fan of your content! Thanks all for these criticisms, we'll definitely take them to heart.

Thank you for the candid feedback! It's definitely appreciated and we'll take note

Ha!! Great question, so this is our fourth round of message testing. We didn't want to A/B directly on HN cause that feels kinda disingenuous, but we've definitely been crunching conversion rates and such.

What sort of data structures are you using for frames and the timeline? Do you have a frame type that you render to some canvas? How do you handle previews etc?

I've been mucking around with Rust and sixtyfps for some animation software, and I am curious how you've handled those basic things.

Best of luck in any case!

Thank you for this list!

The grand vision is a reimagined interface (keeping some familiar things like keyframes and curves, but moving other things forward.) Without getting too deep into the buzzword weeds, I'm hoping to weave in more ML / generative modeling technologies.

I've thought about hacking around in Blender, but it seems a bit overkill for what I'm building - at least as an MVP

I use it to code on my deep learning machine from my macbook. It makes things a bit more secure and ssh-ing is painless that way.

Then I can check experiments from wherever without worrying about a lot of the fiddly details.

I don't see why not. I think they can be used for anything a GNN / Transformer could be used for, but with better probabilistic qualities.

you might find this interesting

https://acme.byu.edu/2021-2022-materials

So, let's say I want to implement these (or other) curves into a drawing program. How do I go about doing that? I mean to say, I want to implement a brush on canvas type interface - how do Bézier curves fit into that?

Do I collect points when mouse is down and build a curve with those points? Do I get some other path information and rasterize?

I'm trying to write my own drawing program (for fun) but given my lack of background in this I feel myself floundering.

This is super cool, next step - full autodiff support to train it in lisp too.

This, unfortunately, resonates quite strongly with me and seems representative of my peers as well

You.com has actually been really good for me.

Hey HN! I'm looking for books on programming curves, key frames, brushes, rigging systems, etc.

My google-foo seems to have failed me here. I've found some reasonable blog posts, but am looking for something slightly more technical and in depth.

And pray we don't alter [the deal] any further

As they say in the start up world "Ship It!"

I've seen lots of advice that says people ship too late. I'm not sure if that's the case here, but something to consider.

If you can, you should consider supporting mobile browsers. This tool could be a powerful way for those with only mobile computing power to improve their situation via coding.

Anecdata here, but every waymo car I've seen here in SF for the past several weeks (probably 30+ cars) has had a safety driver.

The website hijacked the "back" functionality of my browser. Not a fan of that.

Super pedantic comment. GPT-3 has 175 Billion parameters. GPT-2 was the 1.5 Billion model.