For me the page just reloads no matter where I click.

I think most of the comments are talking about the some "virus" but the actual valuable part of the paper is about making choices when interacting with LLMs in such a way that abilities are reversable. That you are still able to do those things without LLMs. And that society and institutions help to preserve this ability by encouraging ways to do things with and without LLMs.

Why not use Codex for that? What is the differnce between work and codex for a very specific task that involves terminals?

Idk, can I do that with Codex on my phone? "Build app to control my specific Soundbar model" and then it just builds it while I'm chilling on the sofa with my phone in the hand?

                       agent memory
                             |
              +--------------+--------------+
              |                             |
       deductive state               episodic memory
              |                             |
       facts / rules / time          fuzzy context
       provenance                    semantic retrieval
       retractions                   source text

Well this turns out, just like human memory functions. (Semantic vs episodic memory)

People are not talking enough how huge this is for robotics and IoT. Current robotics arhitectures are limited by tok/sec. How cares if its not a Fable model?

This move undercuts NVIDIA directly.

With AI models like Mixture of Experts, many of those experts will be the real target here, as polished, refined and little to no change, they become fine candidates for being locked into silicon. Who knows, add some SRAM in there and small changes to those experts could be carried out without needing new silicon.

Maybe AI models may become reduced to a collection of tiles you add to a chips one day, maybe sooner for some areas as you say, motor control for balance, vision systems, speach recognition systems etc, broken down, for robotoics, much is already there and just cost of battery/power holding much back.

An "Expert" is really just an unfortunate name for what amounts to a dense part of a sparse matrix and that's also an oversimplification.

It doesn't actually specialise in anything in particular that one can point to.

For this reason you can really transfer them between models.

Is it still true? I'd assume you should be able to freeze the matrix and unfreeze an expert block, before feeding particularly chosen training data. Or that doesn't work?

That's more or less the idea behind Low-Rank Adaptation, or LoRA.

There's also Mixture of LoRA Experts, which instead of slicing up the model and routing through that, routes through different LoRAs.

But it all comes with tradeoffs, as you have to train and run the gating network doing the routing, which also comes at a cost.

no you're wrong. people are not talking about the load bearing seam that this strong decision has revealed towards veterinary care.

the implications for mental health of pet rats is huge.

I think the one from hugging face is much clearer, albeit its an arena metric and a bit tricky to find the pareto view. Look for the the top Navigation Bar (Agent Chat Code Image Video). Chat -> (dropdown) Text -> (Side panel) View as Pareto.

https://huggingface.co/spaces/lmarena-ai/arena-leaderboard

"The forces that destroyed SA were also what generated its 4×+ return"...yes, that is what levrage means? And it goes both ways.

The interesting part what this article states: SA was essentialy a thematic ETF without any hedging to buffer downside, and got margin called.

Have $300k in disposible income?

add a few grand to get a new breaker panel and a new circuit put in to power it

So I have to Shift + Alt + n, the search for the HN blog / article in the other tab? If you have a lot of tabs this isn't feasable.

Wouldn't it make sense to have it an extension? Just spontainously (without reading any instructions) I don't understand how to use this or what Userscripts Tamper or Violentmonkey is.

I prefer user scripts as I have to install one extension that can load many plugins (user scripts) instead of having one extension for every tiny website tweak.

Also much easier to glance over what a user script is doing vs. unpacking some extension and then finding the actual business logic in all the wrapper code.