vicentwu

6 ms·

1.
by vicentwu 1mo ago
LLMs raise the floor, but you determine the ceiling.
2.
by vicentwu 3mo ago
I want buy one.
3.
by vicentwu 3mo ago
It's absurd. let's mark it down.
4.
by vicentwu 4mo ago
All the ai search platform have to find a way to share value back to content owners, or the whole eco system will collapse.
5.
by vicentwu 4mo ago
Good work! Context-awareness has huge potential. I don't think this demo hit the right mark, but it definitely shed some light.
6.
by vicentwu 6mo ago
Good read.
7.
by vicentwu 7mo ago
cool
8.
by vicentwu 7mo ago
Genuis!
9.
by vicentwu 7mo ago
I like the CRT-like filter effect.
10.
by vicentwu 10mo ago
Hmm... The key is to successfully decompose a big, hard problem into easier atomic sub-problems. However, the decomposition process itself is difficult, and this paper is not about that. They decompose a task using a human-written prompt.
11.
by vicentwu 1y ago
wow....
12.
by vicentwu 1y ago
I like the vibe.
13.
by vicentwu 1y ago
Prompts are really an interesting way of programming, and we can actually express logic containing abstract adjectives like ‘happy’ and ‘unsatisfied’ in a somewhat arbitrary way.
14.
by vicentwu 2y ago
Great!
15.
by vicentwu 2y ago
"As an analogy, imagine that you could put your dog or cat into hibernate mode whenever you left on a trip. Your dog or cat might not notice, but even if they did, they might not mind. Now imagine that you could put your child into hib
16.
by vicentwu 2y ago
RL doesn't need that much static data, it needs a lot of "good" tasks/challenges and computation.
17.
by vicentwu 2y ago
Trained with 300 raw pairs directly from the ARC training set without using any data augmentation process, such as generating many more pairs with some kind of ARC generator? That's amazing.
18.
by vicentwu 2y ago
"Note on "tuned": OpenAI shared they trained the o3 we tested on 75% of the Public Training set. They have not shared more details. We have not yet tested the ARC-untrained model to understand how much of the performance is d
19.
by vicentwu 2y ago
Off the topoc. I think, in the long-term , inference should be done along with some kind of training.
20.
by vicentwu 2y ago
Past efforts leds to today's products. We need to wait to see the real imapct on the ability to ship.
21.
by vicentwu 2y ago
It's amazing!
22.
by vicentwu 2y ago
Fasciating. Language is a type of action evoloved for information exchaging, which maps latent "video", "audio" and "thoughts" into "sentences" and vice versa.
23.
by vicentwu 3y ago
Cool! The real-time feedback will have enormous ramifications on the art creation workflows.
24.
by vicentwu 3y ago
Asking models to do math is kind of an effecitve way to measure their capabilities, especially in reasoning and abstraction, which are quite important for problem solving.
25.
by vicentwu 4y ago
The chain-of-thought works quite like the "System 2" introduced in <<Thinking, Fast and Slow>>, which is slower, more deliberative, and more logical.
26.
by vicentwu 4y ago
I think the last paragraph quite makes sense. It seems "true" that some kind of reasoning capability emerges as LLMs get bigger, which makes those LLMs quite useful and blows a lot of people's minds at the beginning. But, I t
27.
by vicentwu 4y ago
Maybe in the future there might be some kind of "forbidden to use in commercial AI models" policy on websites in the near future, just like what the art community is doing now.
28.
by vicentwu 4y ago
I can't wait to see the model's ability of saying "They don't know", which I think is an important feature if it serves as a search engine, because it can reduce the amount of generated ramblings which it's act
29.
by vicentwu 4y ago
So i think the ability of the search engine to say "I don't know" is very important, and most of current chatgpt like models in the market don't have this feature.
30.
by vicentwu 4y ago
It's fascinating to think about the future landscape of the search and web. Some assumptions: 1. Url-based web will not wither away. 2. Asking questions in the chat-like mode is more natural to people. 3. Generated answers cost more wh

More ›