Ratul Roy

Data Engineer, Data enthusiast. Currently working at @volueinsight

[Ratul Roy @__ratulr__]

Do you remember when you joined X? I do! #MyXAnniversary

[Ratul Roy @__ratulr__]

@cnakazawa Firefox for daily driver, thorium for when ffox does not work

[Tibo @tibo_maker]

I have no idea how I'll teach my kid to recognize AI-generated content 😬

[Ratul Roy @__ratulr__]

@DeeperThrill @rauchg @OpenAI Read THROUGH the whole thread only to find out it’s a vercel marketing campaign with little impacting substance

[Ratul Roy @__ratulr__]

Looking for opinions on Django optimization

What is the best route to refactor a django app that has been calling its own api to get data?

[Ratul Roy @__ratulr__]

I finished 1st place in Obsidian League on @Duolingo!

[Ratul Roy @__ratulr__]

I’m learning Dutch on Duolingo! It’s free, fun, and effective.

[Ratul Roy @__ratulr__]

I’m learning Dutch on Duolingo! It’s free, fun, and effective.

[Ratul Roy @__ratulr__]

I’m learning Dutch on Duolingo! It’s free, fun, and effective.

[Ratul Roy @__ratulr__]

I've just read the Wikipedia article on numeronym and it is *wild* what language-level abstractions we’ve been using

[Ratul Roy @__ratulr__]

this looks promising

[Unsloth AI @UnslothAI]

• Finetune 387% faster TinyLlama
• 600% faster GGUF conversion
• 188% faster DPO
• 400% faster model downloads

All part of the new Unsloth AI release! 🦥

Read our blog: https://t.co/TaR0dcD8we

[Ratul Roy @__ratulr__]

@gabrielfw @hankgreen i guess it's only art when someone is ready to pay 69 million dollars for it

[Ratul Roy @__ratulr__]

@aisinne @openchatdev trick question, there's no such thing

[Ratul Roy @__ratulr__]

#NeurIPS2023 looks fun, maybe I'll attend next year

[Andrej Karpathy @karpathy]

# On the "hallucination problem"

I always struggle a bit with I'm asked about the "hallucination problem" in LLMs. Because, in some sense, hallucination is all LLMs do. They are dream machines.

We direct their dreams with prompts. The prompts start the dream, and based on the LLM's hazy recollection of its training documents, most of the time the result goes someplace useful.

It's only when the dreams go into deemed factually incorrect territory that we label it a "hallucination". It looks like a bug, but it's just the LLM doing what it always does.

At the other end of the extreme consider a search engine. It takes the prompt and just returns one of the most similar "training documents" it has in its database, verbatim. You could say that this search engine has a "creativity problem" - it will never respond with something new. An LLM is 100% dreaming and has the hallucination problem. A search engine is 0% dreaming and has the creativity problem.

All that said, I realize that what people *actually* mean is they don't want an LLM Assistant (a product like ChatGPT etc.) to hallucinate. An LLM Assistant is a lot more complex system than just the LLM itself, even if one is at the heart of it. There are many ways to mitigate hallcuinations in these systems - using Retrieval Augmented Generation (RAG) to more strongly anchor the dreams in real data through in-context learning is maybe the most common one. Disagreements between multiple samples, reflection, verification chains. Decoding uncertainty from activations. Tool use. All an active and very interesting areas of research.

TLDR I know I'm being super pedantic but the LLM has no "hallucination problem". Hallucination is not a bug, it is LLM's greatest feature. The LLM Assistant has a hallucination problem, and we should fix it.

</rant> Okay I feel much better now :)

[Ratul Roy @__ratulr__]

@tiangolo @pydantic Thank you for all your hard work, I was waiting for the support for v2

[Sebastián Ramírez @tiangolo]

Today only, for black Friday! 🎁🖤

🤯 Get @FastAPI with a massive discount of 80% *

You can pay just $0 instead of the regular price of $$$0 🎉

Includes all the easy-to-learn step-by-step tutorials and docs. 🤓

* Today only, or while stocks last.

https://t.co/80OS5r3OkL

[Ratul Roy @__ratulr__]

@marktenenholtz I'm curious abouts how you use polars with/after pandas

What is you choosing criteria for polars over pandas and vice versa?
Memory consumption?

[Hank Green @hankgreen]

We have already willingly given so much power to content recommendation AIs, I think we are underestimating how quickly and even joyfully we will give power to AIs that will know us and talk with us.

The first has been very bad for the fabric of society, I think the second will be worse.

[akhil @fkasummer]

here’s the multiplayer browser engine i’ve been working on in action @braidbrowser

Sound ON + rt

invite link 👇