So when I saw the "About Me" image, I thought Gen AI - https://cookie.engineer/about/me/teaser.jpg But no LLM used for that? (No judgment, just after seeing that image, and then you declaring no LLM use, I thought I'd ask!)
Well, I mean, I self-hosted stable diffusion to be able to generate the teaser images and the award images (for the ranks in the Game Over Dialog).
I only implemented the initial SD paper [1] back then so I have no clue whether the u-net architecture changed when it comes to the semantic mapping of the text/labels. Did that change and is Stable Diffusion 3 now an LLM, actually?
Again this should in no way implicate a diminishing of the work and accomplishment of building your site. Just noticing some kind of "AI" (ANI) was used to make an image. Maybe worth clarifying for pedants like me (but probably not worth it ;-)!)
Instead of kvetching about parent's terminology when you've proven you clearly know what he meant, I suggest the disclaimer:
"No LLMs were used in the making of this website and its content, but self-hosted latent diffusion models were."
> Instead of kvetching about parent's terminology
The point behind knowing how it works vs assuming how it works gives you a difference in understanding and perspective.
I know how it works, because I implemented the papers, and I started way before the current LLM hype. Models like NEAT, HyperNEAT, LSTMs, Bayesian RNNs, GANs, BERT, AutoBERT, AlphaGo are inherently useful if you understand how the model works, what it can do and what it can't. Those tools are great, if you know their purpose and applications.
Post-LLM agents that's a different problem, because a lot of people are assuming it's "AI" that magically does things, while it just hallucinates. So the dangers are higher when it comes to the unawareness of systemic issues and inherent responsibilities of using those tools.
(read also: Attention is all you need, one of the best papers on the topic, even more relevant these days).
PS: I've spent too much effort to comment on a shitposting account already. Anyways, have a great day nonetheless and a Happy New Year!
"I see flat colors and a list of recent project news. Nothing at the bottom. Is it the wrong URL? Or does it not work on my phone? Let's go back and check the HN comments."
"The submission has a bunch of text which doesn't seem to answer my immediate confusion. There's a game somewhere? Was that the wrong link? Did it not pop up correctly? Did I need to give the site special permissions?"
TLDR: This random user unfamiliar with the site is not sure what they were supposed to focus-on and notice/appreciate, and wasn't able to figure it out within a reasonable period before needing to put the phone down again. (Not including the time to write this follow-up comment.)
1. You have your audio off or volume too low.
2. You just don't interact with the bottom cookie prompt.
3. The prompt never shows up for some reason. Saw it a few times, but even after a lot of resetting cookies and reloading, I can't get it to show reliably.
I suggest a special URL parameter so that you can link people to a slightly deeper step in the process where It's obvious what they should be interacting with.
I had to put the initial things in a click listener, and then it's 3 seconds after the initial click when the cookie consent banner is shown (due to AudioContext API limitations and that they have to be a user gesture event that .resume()s the audio context instance).
> I suggest a special URL parameter so that you can link people to a slightly deeper step in the process where It's obvious what they should be interacting with.
The Debugger Views here show more details, and you can play the game just fine and toy around with it:
