Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
GaggiX
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
29 ms
·
691.
▲
by
GaggiX
2y ago
People will learn the language sooner or later, I use Deepl Write less and less. You stopped reading because it's clear from your "anti-tech" blog that you're heavily biased against AI, even when it's used for acces
692.
▲
by
GaggiX
2y ago
There is no contradiction, my logic follows a common approach in skepticism, where claims are treated as false until evidence is provided.
693.
▲
by
GaggiX
2y ago
Deepl and Deepl Write because I'm not a native English speaker.
694.
▲
by
GaggiX
2y ago
No NovelAI for some reason.
695.
▲
by
GaggiX
2y ago
Yes, reality does matter.
696.
▲
by
GaggiX
2y ago
The other comment under mine replied for me. Also, something is false until evidence is provided, not the opposite. In this case, I'm very confident that the post is creative writing.
697.
▲
by
GaggiX
2y ago
>This was something that was bound to happen and is going to happen again Just to clarify, the reddit post is 99% creative writing, it's a fake story created by a new Reddit account.
698.
▲
by
GaggiX
2y ago
This man is probably one of the very few who have read hundreds or possibly thousands of obscure Wikipedia pages about various species and families of moths. Some of the most obscure Wikipedia pages are about species of moth.
699.
▲
by
GaggiX
2y ago
Apple is making sure that it is not easy for Apple users to find out about this fee.
700.
▲
by
GaggiX
2y ago
SAM 2 not only focuses on speed, it actually performs better than SAM (1), the other models instead always trade performance for speed. SAM 2 is able to achieve this result thanks to its Hiera MAE encoder: https://arxiv.org/
701.
▲
by
GaggiX
2y ago
Safety, long range and lifespan are still important to people who charge their EV at home.
702.
▲
by
GaggiX
2y ago
>Why bother remaking it? Reverse engineering the source code makes it much, much easier to mod the game and add more fun stuff.
703.
▲
by
GaggiX
2y ago
The extraction of the eyes is done client side.
704.
▲
by
GaggiX
2y ago
Hacker News turning into Reddit, something that has not been said for more than 15 years now.
705.
▲
by
GaggiX
2y ago
Give me a sec, I will contact Edgar Degas with my telegraph.
706.
▲
The Llama 3 Herd of Models
(arxiv.org)
1 points
by
GaggiX
2y ago
|
0 comments
707.
▲
by
GaggiX
2y ago
The US now controls mathematics, the CIA is so advanced that they are changing how mathematics works to accuse Maduro of fraud, brilliant.
708.
▲
by
GaggiX
2y ago
When you spend most of your funds in a domain, $1.8 million out of a total of $2.5 million.
709.
▲
Xbox console sales continue to crater with 42% revenue drop
(arstechnica.com)
18 points
by
GaggiX
2y ago
|
33 comments
710.
▲
Midjourney v6.1
(midjourney.com)
6 points
by
GaggiX
2y ago
|
0 comments
711.
▲
by
GaggiX
2y ago
Scam, not the actual Suno AI website.
712.
▲
by
GaggiX
2y ago
Some parts seem to compete with Character AI.
713.
▲
by
GaggiX
2y ago
They have a very sci-fi look.
714.
▲
by
GaggiX
2y ago
>The "secret knowledge" that the author describes is no longer a thing. The game shows you all the recipes. It does show the recipes you have unlocked (e.g. by mining a particular one that makes them), I'm glad they did th
715.
▲
by
GaggiX
2y ago
As reported by the article: "However, for smaller matrices, we’re doing poorly in comparison to Nvidia’s library! This happens because cuBLAS contains not one single implementation of SGEMM, but hundreds of them.", it would take c
716.
▲
by
GaggiX
2y ago
This is pretty funny, it seems to work with random foods, iFixit doesn't seem to have anything to do with it.
717.
▲
by
GaggiX
2y ago
>then got crawled by ChatGPT for training Not really, from the example you provided I think it's pretty clear that a custom system prompt was used and ChatGPT is using its own creativity, it doesn't have anything in common with
718.
▲
by
GaggiX
2y ago
>Up until recently, to insert yourself into an image generation algorithm, you had to use a technique like Dreambooth I mean, not really, you could just train a LoRA for example (it doesn't require training with Dreambooth).
719.
▲
by
GaggiX
2y ago
The fact that these models rely heavily on classifier-free guidance has a strong impact on the tones of the image.
720.
▲
by
GaggiX
2y ago
It's a wishlist of the parent comment.
More ›