Live data from Hacker News

Big LLMs weights are a piece of history

antirez.com

51–60 of 222 posts

Re: Big LLMs weights are a piece of history

#52

Just like the map isn't the territory, so summaries are not the content nor the library fillings the actual books. If I want to read a post, a book, a forum, I want to read exactly that, not a simulacrum built by arcane mathematical algorithms.

The counter perspective is that this is not a book, it's an interactive simulation of that era. The model is trained on everything, this means it acts like a mirror of ourselves. I find it fascinating to explore the mind-space it captured.

Re: Big LLMs weights are a piece of history

#53
post #38

Mozilla's llamafile project is designed to enable LLMs to be preserved for historical purposes. They ship the weights and all the necessary software in a deterministic dependency-free single-file executable. If you save your llamafiles, you should be able to run them in fifty years and have the outputs be exactly the same as what you'd get today. Please support Mozilla in their efforts to ensure this special moment i…

LLMs are much easier to port than software. They are just a big blob of numbers and a few math operations.

Re: Big LLMs weights are a piece of history

#54
post #2

That's really what these are: something analogous to JPEG for language, and queryable in natural language. Tangent: I was thinking the other day: these are not AI in the sense that they are not primarily intelligence . I still don't see much evidence of that. What they do give me is superhuman memory. The main thing I use them for is search, research, and a "rubber duck" that talks back, and it's like having an inter…

> JPEG for [a body of] language

Yes!

> artificial memory

Well, "yes", kind of.

> Memex

After a flood?! Not really. Vannevar Bush - As we may think - http://web.mit.edu/STS.035/www/PDFs/think.pdf

Re: Big LLMs weights are a piece of history

#55
post #2

That's really what these are: something analogous to JPEG for language, and queryable in natural language. Tangent: I was thinking the other day: these are not AI in the sense that they are not primarily intelligence . I still don't see much evidence of that. What they do give me is superhuman memory. The main thing I use them for is search, research, and a "rubber duck" that talks back, and it's like having an inter…

I can ask a LLM to write a haiku about the loss function of Stable Diffusion. Or I can have it do zero shot translation, between a pair of languages not covered in the training set. Can your "language JPEG" do that?

I think "it's just compression" and "it's just parroting" are flawed metaphors. Especially when the model was trained with RLHF and RL/reasoning. Maybe a better metaphor is "LLM is like a piano, I play the keyboard and it makes 'music'". Or maybe it's a bycicle, I push the pedals and it takes me where I point it.

Re: Big LLMs weights are a piece of history

#57

I love the title "Big LLMs" because it means that we are now making a distinction between big LLMs and minute LLMs and maybe medium LLMs. I'd like to propose the we call them "Tall LLMs", "Grande LLMs", and "Venti LLMs" just to be precise.

LLM, LLM 2.0, LLM 3.0, Mini LLM, Micro LLM, LLM C.

LLM 95, LLM 98, LLM Millennium Edition, LLM NT, LLM XP, LLM 2000, LLM 7

I really appreciated the way they managed to come up with a new naming scheme each time, usually used exactly once.

Re: Big LLMs weights are a piece of history

#58
post #23

I love the title "Big LLMs" because it means that we are now making a distinction between big LLMs and minute LLMs and maybe medium LLMs. I'd like to propose the we call them "Tall LLMs", "Grande LLMs", and "Venti LLMs" just to be precise.

Big LLM is too long as a name. We should agree on calling them BLLMs. Surely everyone is going to remember what the letters stand for.

Bureau of Large Land Management

Re: Big LLMs weights are a piece of history

#59

I love the title "Big LLMs" because it means that we are now making a distinction between big LLMs and minute LLMs and maybe medium LLMs. I'd like to propose the we call them "Tall LLMs", "Grande LLMs", and "Venti LLMs" just to be precise.

Why not LLLM for large LLM’s and SLLM for small LLM’s, assuming there is no middle ground

What makes it a Small Large Language Model? Why jot just an SLM?

Re: Big LLMs weights are a piece of history

#60

Earlier quoted context omitted.

Why not LLLM for large LLM’s and SLLM for small LLM’s, assuming there is no middle ground

What makes it a Small Large Language Model? Why jot just an SLM?

If we can’t have fun with names, why even be in IT?
Post reply on HN