Lee

Ζωή μου, σᾶς ἀγαπῶ.

[Lee @twphoenix]

@NaturalWorld515 The sign at the bottom left says, in simplified Chinese, "Take care of the lawn."

[Lee @twphoenix]

@thewitcher @Fools_Theory Oh my Yennefer!

[Lee @twphoenix]

@Nature Nature & Rubbish. Gods in the heaven & drudges in the dust.

[Lee @twphoenix]

@xinwendiaocha A  little boy crying his eyes out, shaking like a leaf while covering his face, screaming his heart out: "It's all lies! None of this is real! Everything I see is fake!"
Dude, you can't even look your opponent in the eye anymore.

[Lee @twphoenix]

@clashreport Sigh.

[Lee @twphoenix]

@DiagenosRealist @ShangguanJiewen One interesting thing, that's a new story told by US media.

[International Defence Analysis @Defence_IDA]

The Red Wings Striker combat robot dog, developed by China, represents a breakthrough in unmanned warfare. Equipped with micro rockets

[Lee @twphoenix]

@clashreport

[Lee @twphoenix]

@Lloydlucianvert @creepydotorg smart!

[Lee @twphoenix]

@creepydotorg Keep away!

[Lee @twphoenix]

My working paper: "The Collective Dream: Large Language Models as Reflections of Human Collective Memory and Dreamlike Information Processing"

[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 :)

[Lee @twphoenix]

@karpathy The current LLM is trained based on the literal "memory" of all human beings (including fictitious memory), and due to the lack of a mechanism similar to the "prefrontal lobe", all current AI based on LLM , seems to have become a "collective dream" for all mankind.

[Shunsuke Saito @psyth91]

📢 Check out 𝗥𝗲𝗹𝗶𝗴𝗵𝘁𝗮𝗯𝗹𝗲 𝗚𝗮𝘂𝘀𝘀𝗶𝗮𝗻 𝗖𝗼𝗱𝗲𝗰 𝗔𝘃𝗮𝘁𝗮𝗿𝘀!
Our latest codec avatars using 3D Gaussians generalize to novel lighting (OLAT, envmap) with *all-frequency* reflection (see video for hair and eye reflection) in real-time! https://t.co/hmO0she2t6

[Lee @twphoenix]

@psyth91 Impressive!

[Lee @twphoenix]

Human beings can only decide when a war will break out, but they cannot determine when a war will end or how it will end.

[Lee @twphoenix]

@elonmusk Hints 1: 2022 is the Chinese tiger year. Hints 2: the old Chinese saying, "Don't pat a tiger on the ass."

[Lee @twphoenix]

@CogSec A new series!
How to cite your work?