hughzhang

5 ms·

1.
Reconstructing OpenAI's O1 test-time scaling law graphs(github.com)
2 points by hughzhang 2y ago|0 comments
2.
Planning in Natural Language Improves LLM Search for Code Generation(arxiv.org)
3 points by hughzhang 2y ago|0 comments
3.
Planning in Natural Language Improves LLM Search for Code Generation(arxiv.org)
3 points by hughzhang 2y ago|0 comments
4.
Chain-of-Thought Reasoning Is a Policy Improvement Operator(arxiv.org)
2 points by hughzhang 3y ago|0 comments
5.
Why transformative AI is hard to achieve(zhengdongwang.com)
1 points by hughzhang 3y ago|0 comments
6.
Why transformative artificial intelligence is hard to achieve(thegradientpub.substack.com)
1 points by hughzhang 3y ago|0 comments
7.
Gradient Turns 3(thegradientpub.substack.com)
1 points by hughzhang 5y ago|0 comments
8.
Transformers Are Graph Neural Networks(thegradient.pub)
132 points by hughzhang 6y ago|25 comments
9.
Quantifying Independently Reproducible Machine Learning(thegradient.pub)
2 points by hughzhang 7y ago|0 comments
10.
Is NeurIPS getting too big?(thegradient.pub)
4 points by hughzhang 7y ago|0 comments
11.
An Epidemic of AI Misinformation(thegradient.pub)
125 points by hughzhang 7y ago|59 comments
12.
Artificial Life for AI People(thegradient.pub)
37 points by hughzhang 7y ago|2 comments
13.
Machine Learning Can Help Unlock the World of Ancient Japan(thegradient.pub)
105 points by hughzhang 7y ago|8 comments
14.
Gaussian Processes, Not Quite for Dummies(thegradient.pub)
3 points by hughzhang 7y ago|0 comments
15.
Evaluation Metrics for Language Modeling(thegradient.pub)
3 points by hughzhang 7y ago|0 comments
16.
The State of Machine Learning Frameworks(thegradient.pub)
752 points by hughzhang 7y ago|193 comments
17.
by hughzhang 7y ago
Sometimes people idealize academia though. It's important to realize that everything has its flaws :).
18.
BenderRule: Name the Language You Study(thegradient.pub)
2 points by hughzhang 7y ago|1 comments
19.
NLP's Clever Hans Moment Has Arrived(thegradient.pub)
3 points by hughzhang 7y ago|0 comments
20.
New NeurIPS Workshop: ML Retrospectives(thegradient.pub)
1 points by hughzhang 7y ago|0 comments
21.
Research Retrospectives(thegradient.pub)
2 points by hughzhang 7y ago|0 comments
22.
We Open Sourced Grover(thegradient.pub)
3 points by hughzhang 7y ago|0 comments
23.
The Past, Present, and Future of AI Art(thegradient.pub)
4 points by hughzhang 7y ago|0 comments
24.
Goodhart’s Law: Are Academic Metrics Being Gamed?(thegradient.pub)
3 points by hughzhang 7y ago|2 comments
25.
Teaching AI to Plan Ahead(thegradient.pub)
2 points by hughzhang 7y ago|0 comments
26.
The Promise of Hierarchical Reinforcement Learning(thegradient.pub)
7 points by hughzhang 8y ago|0 comments
27.
OpenAI Shouldn’t Release Their Full Language Model(thegradient.pub)
3 points by hughzhang 8y ago|0 comments
28.
by hughzhang 8y ago
But still much less dangerous than stuff like nuclear fallout or orwellian propaganda? This isn't to say that unrealistic beauty standards aren't a problem, just one that isn't worth trying to stop the march of technology for
29.
by hughzhang 8y ago
Thanks for your comments! On 1) photo manipulation. If "dangerous" means that random, relatively harmless fake facts easily spread, I'm inclined to agree. I'm sure I've been tricked by a whole host of random facts w
30.
Dear OpenAI: Please Open Source Your Language Model(thegradient.pub)
289 points by hughzhang 8y ago|124 comments

More ›