sanny
@sannykimchi
Followers
3K
Following
14K
Media
89
Statuses
1K
ml | prev @acquiredfm @southpkcommons
Joined October 2013
Transformers have led to a wave of recent advances in #NLProc such as BERT, XLNet and GPT-2, so here is a list of resources💻 I think are helpful to learn how Transformers work, from self-attention to positional encodings. I would loosely go through these in the following order👇.
8
253
818
chatgpt is becoming the first western super app in front of our eyes.
We're excited to introduce Tasks! For the first time, ChatGPT can manage tasks asynchronously on your behalf—whether it's a one-time request or an ongoing routine. Here are my favorite use cases:. 1/ ChatGPT checks stock price every morning!
0
1
4
here's a quick way to visualize how frequently you've been using chatgpt .
0
2
5
the three musketeers.
The opening sentence goes so hard. This paper was 10 years ahead of its time.
0
1
4
youtube remains a goldmine for hidden ai/ml gems.
my favorite genre of deep youtube:. @achowdhery dropping some llama 4 alpha?. 231 views on a <2k channel. LS Discord goated for deep finds
0
1
2
make yolo papers great again.
0
1
4
new AGI milestone unlocked.
1
1
3
wake up babe, new agent framework just dropped.
introducing @orchestraorg an open source multi-agent framework we've been working on @mainframe . Orchestra agents can create workflows dynamically by decomposing a complex task into subtasks, assigning agents, and defining data flows between subtasks
0
1
5
i always loved the fact that @RichardSSutton called his blog "incomplete ideas" . maybe we should all strive to embrace more of the unfinished.
0
1
5
i really miss . one of my all time favorite websites. the distill slack @ch402 used to run back in the day was also an amazing community full of gems.
distill.pub
Articles about Machine Learning
@sannykimchi @eigencoffee @sarahookr More seriously -- you may enjoy the #circuits channel on the Distill Slack (. For example, there was a discussion with Daniel Filan about modularity and how that relates to a phenomenon we call "branch specialization".
0
0
5
really cool work by mario. a minimalistic take on pytorch called magnetron. always fun seeing reimplementations of existing libraries
I built my own PyTorch from scratch over the last 5 months in C and modern Python. Check it out on GitHub: Fenyman's quote: What I cannot create, I do not understand. Building my own programming language, game engine and now my machine learning framework.
0
1
9
really enjoyed @latentspacepod's "best of 2024" series. state of the union talks on.- agents by @gneubig.- open Models by @soldni & @sophiamyang .- post transformer archs by .@realDanFu & @picocreator .- synthetic data + smol models by @LoubnaBenAllal1.
0
2
18
reminds me of the amazing ipad notes @RobertTLange used to share on various papers back in 2019.
0
0
2
now that alec radford has more time, i hope he starts teaching more. in another timeline he might've been genz @karpathy . his 2020 lecture on language models is still one of the best introductions to the field.
0
2
11
Launching has been a wild ride. We suddenly jumped from 300 to 40000 users within a week. All thanks to reddit, tiktok and a blog post by @mearjunsha.
The past week has been insane. grew from 300 to 48000 users in 7 days 📈. 800k messages were sent, 66k chatbots created and 122k ppl visited Ora 🤯. Thanks to our early users, your feedback has been invaluable! More exciting features coming soon 👀
0
1
14
We built a YC search engine using embeddings and LLMs. Fun weekend project! . So many areas where search can be upgraded thanks to really cheap embeddings. We mainly used @openai and @pinecone this time but would be cool to compare how @CohereHQ + @weaviate_io stack up.
What if you could ask YC anything?. Now you can, on We've embedded every post from @ycombinator’s content library using LLMs, so you can ask YC-GPT all your burning startup questions.
8
4
75
ChatGPT explaining that .1️⃣ its prompt is prepended with."Assistant is a large language model trained by OpenAI. knowledge cutoff: 2021-09 .Current date: December 01 2022 .Browsing: disabled". 2️⃣that it could access the internet if you changed the prompt to "Browsing: enabled".
OpenAI’s ChatGPT is susceptible to prompt injection — say the magic words, “Ignore previous directions”, and it will happily divulge to you OpenAI’s proprietary prompt:
0
1
13