Meta Superintelligence Labs' first paper is about RAG

Also, that wasn't based on purely hearsay, Zuck explicitly said:

> We believe the benefits of superintelligence should be shared with the world as broadly as possible. That said, superintelligence will raise novel safety concerns. We'll need to be rigorous about mitigating these risks and careful about what we choose to open source. Still, we believe that building a free society requires that we aim to empower people as much as possible. [0]

[0]: https://www.meta.com/superintelligence/

MSL is not only those few high profile hires.

I can't think of it ever really paying off. Bell Labs is the best example. Amazing research that was unrelated to the core business off the parent company. Microsoft Research is another great one. Lots of interesting research that .. got MS some nerd points? But has materialized into very very few actual products and revenue streams. Moving AI research doesn't help Meta build any motes or revenue streams. It just progresses our collective knowledge.

On the "human progress" scale it's fantastic to put lots of smart people in a room and let them do their thing. But from a business perspective it seems to almost never pay off. Waiting on the irrational charity of businesses executive is probably not the best way to structure thing.

I'd tell them to go become academics.. but all the academics I know are just busy herding their students and attending meetings

This is very true, and more than just in ai.

I think if they weren’t so metric focused they probably wouldn’t have hit so much bad publicity and scandal too.

Quite the statement for anybody who follows developments (without excluding xAI).

Well for starters you need a leader who can rally the troops who "think(s) different" - something like a S Jobs.

That person doesnt seem to exist in the industry right now.

Doesn't really scream CEO of AGI to me.

Throwing everything into one large context window is often impractical - it takes much more time to process, and many models struggle to find information accurately if too much is going on in the context window ("lost in the middle").

The "classic" RAG still has its place when you want low latency (or you're limited by VRAM) and the results are already good enough.

My impression is that GPT-5 gets confused, not quite right away, but after a couple of pages it has no idea. It doesn't take pages on pages before it forgets things.

In both cases for "Question Answering" it's about similarity search but there are two main orthogonal differences between RAG and Non-RAG :

-Knowing the question at the time of index building

-Higher order features : the ability to compare fetched documents with one another and refine the question

Non-RAG, aka multi-layer (non-causal) transformer with infinite context, is the more generic version, fully differentiable meaning you can use machine learning to learn how to Non-RAG better. Each layer of the transformer can use the previous layer to reason and refine the similarity search. (A causal transformer know the question at the time when it is feed the question, and can choose to focus it's attention on different part of the previously computed features of the provided documents but may benefit from having some reflection token, or better : be given the question before being presented the documents (provided you've trained it to answer it like that).)

RAG is an approximation of the generic case to make it faster and cheaper. Usually it breaks end-to-end differentiability by using external tools, so this mean that if you want to use machine learning to learn how to RAG better you will need to use some variant of Reinforcement Learning which is slower to learn things. RAG usually don't know the question at the time of index building, and documents are treated independently of each other, so no (automatic) higher order features (embeddings are fixed).

A third usual approximation, is to feed the output of RAG into Non-RAG, to hopefully get the best of both world. You can learn the Non-RAG given RAG with machine learning (if you train it with some conversations where it used RAG), but the RAG part won't improve by itself.

Non-RAG need to learn so it needs a big training dataset, but fortunately it can pick-up question answer pair in an unsupervised fashion when you feed it the whole web, and you only need a small instruction training and preference optimization dataset to shape it to your need. If performance isn't what you expect in a specific case, you can provide more specific examples and retrain the model until it gets it and you get better performance for the case you were interested in. You can improve the best case but it's hard to improve the worst case.

RAG has more control on what you feed it but content should be in a more structured way. You can prevent worst cases more easily but it's hard to improve good case.

RAG is confusing, because if you look at the words making up the acronym RAG, it seems like it could be either of the things you mentioned. But it originally referred to a specific technique of embeddings + vector search - this was the way it was used in the ML article that defined the term, and this is the way most people in the industry actually use the term.\

It annoys me, because I think it should refer to all techniques of augmenting, but in practice it's often not used that way.

There are reasons that specifically make the "embeddings" idea special - namely, it's a relatively new technique that actually fits LLM very well, because it's a semantic search - meaning, it works on "the same input" as LLMs do, which is a free-text query. (As opposed to a traditional lookups that work on keyword search or similar.)

As for whether RAG is dead - if you mean specifically vector-embeddings and semantic search, it's possible - because you could theoretically use other techniques for augmentation, e.g. an agent that understands a user question about a codebase and uses grep/find/etc to look for the information, or composes a search to search the internet for something. But it's definitely not going to die in that second sense of "we need some way to augment LLMs knowledge before text generation", that will probably always be relevant, as you say.

Of course here we are talking about a lab, not an individual person, but still I haven't heard of first papers being considered special in any way, even for labs.

(preferably using representative language from the article)

The fact that dot product addition can encode the concept of royalty and gender (among all other sorts) is kind of magic to me.

But similar ways to reduce huge numbers of dimensions to a much smaller set of "interesting" dimensions have been known for a long time.

Examples include principal component analysis/single value decomposition, which was the first big breakthrough in face recognition (in the early 90s), and also used in latent semantic indexing, the Netflix prize, and a large pile of other things. And the underlying technique was invented in 1901.

Dimensionality reduction is cool, and vector embedding is definitely an interesting way to do it (at significant computational cost).

Which means that modifications to the architecture, and combining it with other components and approaches, are the next likely step. This paper fits that.

"Send this through the math coprocessor." "Validate against the checklist." "Call out to an agent for X." "Recheck against input stream Y." And so on.

Retrieval augmentation is only one of many uses for this. If this winds up with better integration with agents, it is very possible that the whole is more than the sum of its parts.

It's effectively a multimodal model, which handles "concept" tokens alongside "language" tokens and "image" tokens.

A really big conceptual step, actually, IMO.

Which other under pressure labs are you talking about?

2. Wild claim that the companies that sell LLMs are actually downplaying their capabilities instead of hyping them

https://www.infoworld.com/article/4061078/the-productivity-p...