shubham13596

6 ms·

1.
It's not a lie if you believe it: LLMs defend their most fluent memory(shubhamg.bearblog.dev)
2 points by shubham13596 2mo ago|0 comments
2.
Notes on RLHF Book by Nathan Lambert(shubhamg.bearblog.dev)
1 points by shubham13596 9mo ago|0 comments
3.
by shubham13596 10mo ago
There are many language-learning apps, but almost none that focus on improving conversational Hindi for kids. Made this web app for my nephew, based in Singapore, after watching him struggle to find anyone to practice Hindi with outside of
4.
Show HN: Conversational Hindi tutor for Indian diaspora kids (5-9yrs old)(hindispeakingtutor.in)
2 points by shubham13596 10mo ago|0 comments
5.
Indic NLP: AI4Bharat papers explainer part 1(shubhamg.bearblog.dev)
1 points by shubham13596 1y ago|0 comments
6.
Humans as Bottleneck(shubhamg.bearblog.dev)
1 points by shubham13596 1y ago|0 comments
7.
by shubham13596 1y ago
this is interesting but anyone has tried to create a small scale version of their own search engine using only their bookmark links. I am surprised why that's not more common. If someone has 5k+ bookmark links and wants only results fr
8.
by shubham13596 1y ago
Yes I think having an LLM generate such OnlyConnect style questions (with the right prompting) should solve for the problem this benchmark seems to have of LLMs, most likely, being trained on past years OnlyConnect questions.
9.
by shubham13596 2y ago
Thanks for views. I can understand that such a tool might not be relevant for someone with your reading style. However, this is no manipulative algorithm. It is just a tool to help you decide your next read based on your own reading history
10.
Show HN: Decide your next read from any booklist basis your Goodreads history(bookrecommendations1-hcr77cz3ga-el.a.run.app)
2 points by shubham13596 2y ago|2 comments
11.
by shubham13596 2y ago
Why not use GPT 4 embeddings instead of word2vec ? Won't that be more effective
12.
by shubham13596 2y ago
Very good resource on building from the basics in a concise manner.
13.
by shubham13596 4y ago
I want it to be more like a proof of concept. Building a scalable and fully usable long term will require a lot of investment which I don't think would be the right thing to do until I have some validation from users. Does that help?
14.
Ask HN: How can non-developers build products/MVPs?
2 points by shubham13596 4y ago|3 comments