I wonder if a kernel MMD, as an efficiently estimated probability metric, would work in this regard

Unfortunately the Wasserstein metric is rather computationally tough to just apply to LLMs

Don't you already have impostor syndrome

Is there a demand for a slice of Apple?

Better conclusion: Our brains keep growing all through our life, it's just that when we're old other processes kick in that get in the way of that

Newsom has the chance to do something awesome in this regard

Those nuclear blocklists have got to be kept in sync somehow

Who can I ping to ask for technical details about bsky

I wonder if it becomes more feasible if you incorporate yourself as part of the network so you can watch for things as they pass through your node

Your other posts are fine. You've made up for your mistake, don't let it stop you permanently.

The Chinese response to needing more energy:

www.reuters.com/business/ene...

The reason he posed it that way is currently on display:

There is a significant group of people on this site who imagine that brigading replies and writing abusive quote-posts will do something to change policy.

Or if they don’t really imagine that, at least it’s how they get off.

Transformers are still unbeaten at finding associations between tokens over long distances

They're gradually working their way up to a full state-space model, pending research

Kimi uses a 3/4 state space model right now (the other 1/4 are the good ol transformers)

Relevant research

bsky.app/profile/timk...

Also select them randomly from a pool, and give them some compensation for their judgement

You know, what if we gathered 12 of them together to decide cases?