byyoung3

7 ms·

31.
by byyoung3 1y ago
Bruh
32.
by byyoung3 1y ago
This is the problem with America
33.
by byyoung3 1y ago
until you get melanoma
34.
by byyoung3 1y ago
isnt the h20 nerfed anyways? H20’s FP16/BF16 performance is reduced to ~148 TFLOPS vs ~1,979 TFLOPS for the H100?
35.
Albania Appoints AI Minister Agent for Public Procurement(wandb.ai)
3 points by byyoung3 1y ago|0 comments
36.
X and XAI Sue Apple and OpenAI over Alleged AI Monopoly(wandb.ai)
2 points by byyoung3 1y ago|0 comments
37.
by byyoung3 1y ago
^^^^^^^
38.
by byyoung3 1y ago
"there's nothing in here that indicates they have frozen hiring or that Zuckerberg is cautious of a bubble" That's what makes it clickbait my friend
39.
by byyoung3 1y ago
clickbait. read the article. they just spent several billion hiring a leadership team. They are doing an all hands to figure out what they need to do.
40.
by byyoung3 1y ago
this seems to disagree with a lot of research showing RL is not necessary for reasoning -- im not sure about alignment
41.
by byyoung3 1y ago
stupid
42.
by byyoung3 1y ago
inside chatgpt
43.
by byyoung3 1y ago
they aren't downplaying anything.
44.
by byyoung3 1y ago
hahahahahahahahhahhahha it's a marginal improvement.
45.
by byyoung3 1y ago
great work
46.
by byyoung3 1y ago
then why convert to for-profit?
47.
by byyoung3 1y ago
i think that this is a bit of an exageration, but i see what you are saying. Anything more than 4-5 re-prompts is diminishing.
48.
by byyoung3 1y ago
I believe is AI. A programmer is now 3x productive. As a ceo, you realize you can probably let a few people go without consequences. Also, CEO’s look and see what Elon did at twitter, and think maybe they can do the same. I wouldn’t underes
49.
O4-mini vs. Claude 3.7 vs. Gemini 2.5 Pro on code generation(wandb.ai)
2 points by byyoung3 1y ago|1 comments
50.
by byyoung3 1y ago
YC mainly trying to make a buck, I'm weary of these sorts of lawsuits
51.
Hugging Face launches "Open Computer Agent"(wandb.ai)
2 points by byyoung3 1y ago|0 comments
52.
Fine-Tuning Qwen3(wandb.ai)
5 points by byyoung3 1y ago|0 comments
53.
by byyoung3 1y ago
its an if statement on whether the model has downloaded or not
54.
by byyoung3 2y ago
SOTA performance. the LiDAR ones have basically failed or completely overfit to a single city.
55.
by byyoung3 2y ago
now they just have to figure out how to make it charge while people are lighting them on fire
56.
by byyoung3 2y ago
You might not like Elon but that doesn’t make him wrong.
57.
by byyoung3 2y ago
This is valid but also hard to back up with any alternatives. At the end of the day it’s just a neural network with backprop. New architectures will likely only be marginally better. So either we add new algorithms on top of it like RL, cre
58.
by byyoung3 2y ago
It’s not new and only superior in a very narrow set of categories.
59.
by byyoung3 2y ago
99.9999 pct sure it's different
60.
by byyoung3 2y ago
I think it might be a trick to get more people to use 4.5

More ›