Theory42

Yeah! LLM compression is a spectrum between 'normal stuff we can read' and vectors. Depending on trust in the models and the desire for compression, there's a choice to be made on which formats you want to allow. Pretty interesting stuff.

A good point. I initially had a model for a judge, but it seemed to give very lenient scores. I'm open to learning about how best to benchmark the phenomenon, though.

I wrote it with the help of an LLM, then edited it, then wrote some more, then edited that. Took about a week to format. It contains my thoughts. Brave new world, I know. I'm finding it ironic that the same crowd that is merging AI code all day is so allergic to AI assistance in prose...do all your work using this new godlike technology, but when it comes time to share, whip out your fountain pen or else.

A useful critique, thanks.

I would say that the grader has the same threshold for whatever answer it receives, and is equally harsh on whichever it grades. Any scores above the baseline (1.0) are really claims about parity, rather than better understanding in the compressed format.

The decoder step is a model expanding the cablese to regular text, not having seen the initial question. A separate model instance then reads that regular text and answers. And the result is still at parity with the plaintext record.

Had cablese knocked out information, that wouldn't have been the result, would it?

I use the tools I study, nothing more or less.

Cool story bro. Maybe you could engage with the content? I'm an actual person.

I suspect you might be right. My conclusions from the digging are that this kind of compression works best with settled instructions/data for machine to machine talk. For something like OpenClaw (which I use a lot), that might mean the AGENTS.md, TOOLS.md, etc. Compression there would free up the context for the agent.

Not quite, and its addressed in the post. Caveman was cool, but hand wavy. I've measured the compression of Cablese, determined where it does the most good, and also that it is already baked in to most model family's training, which saves instruction tokens.

You'd think they could give the scans to archive.org or similar. Even if the books are in copyright, they won't be forever. We can't let our culture be destroyed like this.

Detecting genuine human angst in the sea of AI posts

Say it ain't so

I really liked this talk as well. One part that I'm not sure I can fully agree with, though, is the idea of fully re-conceptualizing the self. It is possible to self-author, and partially change...but I have never heard of nor met anyone who just became a totally different person. I'm willing to concede that he may have been speaking hypothetically, or that maybe the idea of changing the self will be more accessible to AGI rather than for humanity.