Anchit Shrivastava

AI Lead @ Joget and almost a Data Scientist

[Anchit Shrivastava @anchit2000]

@imaachman Much awaited!

[Anchit Shrivastava @anchit2000]

@juliali @photonhq @danieltian link

[Anchit Shrivastava @anchit2000]

@IamEmily2050 YAML over JSON? 🤔
Historically, I’ve seen YAML performing better than JSON, since the tokenizer can focus on what’s important and not the extra “{“ “}” tokens.

[Anchit Shrivastava @anchit2000]

@jeremyphoward Have you checked out Django Unfold?

[Anchit Shrivastava @anchit2000]

@marktenenholtz 1. Lack of inherent temporal ordering
2. Limited local context
3. Lack of domain inductive biases

[Anchit Shrivastava @anchit2000]

@ravithejads @sumanthd17 LLMs are trained on text data for generating text data.
Even though LLMs have gotten quite good at reasoning, we should not forget that the reasoning is just a hallucinated probability distribution of words.
Rather than having LLMs generate scores we can generate categories.

[Anchit Shrivastava @anchit2000]

Karpathy said that hallucinations is all LLMs do.
It is only when it is wrong we begin to have problems.

Don’t us humans do the same?

[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 :)

[Anchit Shrivastava @anchit2000]

@_jasonwei @hwchung27 I had published a paper back in 2021 if anyone wants to get a deeper insight of this method.

https://t.co/GmQ5Bc5V90

[Anchit Shrivastava @anchit2000]

@tunguz Small world graphs 🤔

[Anchit Shrivastava @anchit2000]

@marktenenholtz And let me tell you, $ is the solution to wrong lookup values (iykyk) 😂😂

[Anchit Shrivastava @anchit2000]

@marktenenholtz Even I learnt it last week when my boss felt pity of me manually pasting values 💀😂