https://chatjimmy.ai/ blew my mind at how fast etched model weights can be.

For on-device LLMs, there's a point of diminishing returns, meaning you don't need to have the latest frontier model for most operations.

I think in 2008 I walked by it and said "oh, that's a funny art installation." Then the phone rang, and I said "wait, what? is that a working pay phone?"

I picked it up and someone asked me if I was satisfied with my home insurance.

You could spend the entire week high as a kite, biking around and just looking at everything without talking to anyone, though the vast majority of the community are people you probably would enjoy interacting with. I find most burners thoughtful, compassionate, playful, caring, etc.

What I suspect would happen is you'd start out this way ^, but then you'd gradually discover that the community is great and interactions with people are pretty fun actually. Part of the experience is the opportunity to get outside of your normal baseline personality if you want. You can be anyone for a week. Most people don't even want to talk about their lives off the playa.

Culture changes on the weekend, when you get a lot of "partiers" just for the man and temple burn. There are a bit more irresponsible and inconsiderate people then, but not enough to ruin the overall community vibe.

They can take it for a "reasonable amount of time" (inconvenience you for a few hours and make you miss your connecting flight) while they copy an encrypted image. They then must return it to you.

This administration attacks any body or authority that has the power to hold them accountable. See: attacks on the judiciary, attacks on the media, attacks on academia, etc.

https://pubmed.ncbi.nlm.nih.gov/25271984/

In 88% (36 out of 41) of patients with high-risk HPV, taking reishi + turkey tail mushrooms daily for 2 months cleared the virus, versus 5% clearance (1 out of 20) in the control group. I suspect that 88% clearance rate would be higher if they continued for longer than 2 months.

I think you're saying that you can eventually train models to arrive at that destination by training on existing jokes, effectively encoding these leaps as probabilities.

In that case, the model isn't actually making an intuitive/comedic leap; they're just following new probability chains in attempting to approximate examples they've seen in training.

I'm suggesting that something architecturally different is necessary to create a model which can make intuitive/comedic leaps.

Try to get a frontier model to write a clever, funny joke which hasn't been seen before. Or, try to get it to make an intuitive leap that leads to a novel discovery.

You can use them to guide your own efforts along these lines, as a sounding board. But with current architecture I just don't think either is possible for an LLM to do on its own.

LLMs are inherently probabilistic, and there's currently no mechanism for producing an orthogonal directional change in the path traced through a latent space which is also contextually relevant (landing on a punch line).

In other words, LLMs are fundamentally incapable of making intuitive/orthogonal leaps in context.

It might be possible to add this capability with a new architectural component like transformers, but specifically for making "left turns"/intuitive leaps.

[1] https://wnmurphy.com/llms-cant-do-humor/

However, if you're serious about it, then why not hand-write your LLM-generated code? You'll get much better retention and comprehension than typing.

Think of it like this: if you can offload a task to an AI model that requires reasoning or representing a nonlinear system, you are offloading intelligence.

You're using available information (copyrighted works, or the output of another model) to train a model to encode the information in a new form. Why is the former not theft, but the latter is theft?