gnarlouse

Anybody can ELI5?

Honestly even just a bookmarking system would go a long way to fix the problem, but getting "off-topic" in LLMs (3 years later and trillions of dollars of investment) still ruins my flow entirely. I end up playing an unwanted game of scrollbar warrior jumping back and forth between two points in chat. Could be that I'm studying a textbook and looking to enhance/check my understanding with Claude, or that I'm working on complicated tasks in my codebase. Side tangents always crop up. LLMs don't handle it well. Blows my mind that it doesn't bother anybody else, perhaps IATAH

The best example that comes to mind is orthogonal "thread pulling," and the fact that it's not immediately evident that you'd want to do it like this, makes me wonder if I'm the weirdo.

Say you're studying something: STEM topics, Liberal arts topics, a codebase, a textbook, etc XYZ. You're in the flow of discussing a main topic. Invariably, the main topic is built on top of atomic subelements. For example, discussions about Bonds invariably invoke discussions of Present Value, Yield to Maturity.

The branching is because you have a side query that is a potentially unbounded stack of "wait, but why this?" You highlight a segment of text in the current chat, click a button that pops a new conversation over to the right, and continue your "related but off-topic question." You have both context threads in view at once: the main topic & the side topic. You should be able to do this N layers deep.

The alternative is that you get to play scrollbar warrior in the current regime of LLM-designs. It's not really conducive to deep understanding.

This kind of conversation applies naturally to software development too: "wait, so we're working on feature X/refactor X/bugfix X, but that requires us to answer this subthread. And sure these days Claude Fable can probably do it all with the right harness, but does that mean you _understand_ what it just did?

@julesrms thank you for lending more credibility to the idea that agent conversations need to be natively branching. It's crazy that we're almost 3 years in on this LLM joyride and we still don't have the ability to do, like, reddit/slack-style "aside" convo threads.

~~edit~~

wait, are you doing that? Love JUCE btw

I've sort of been thinking about this in terms of software dev lifecycle & PRs/code review.

Maybe PRs these days need to "test" the human? "Explain how this code works, without AI, or it gets rejected."

What the AI post

I'm still "trying to make 'fetch' happen" with the whole "billionaire alignment problem" thing. Maybe the phrase will catch on.

Is the intent to replace the IP protocol ever?

We need a new social media platform purely for Apple product experiences. Stay with me. People post their experiences with various parts of all their products, from hardware button position to software design and behavior. Upvotes are "It's Genius", downvotes are "It's Shit" -- because Apple has completely shirked its much needed Jobsian specter.

The joke, of course, is that I imagine a good 75% of the reviews would be "it's shit."

Superintelligence Alignment conversations are a misdirect.

Let's talk about Billionaire Alignment, Economic alignment, Human alignment.

Classware should be M.A.D. -- in that it shouldnt even happen.

good lord

Spidermouth the Night Clown

Thanks for my next horror shortfilm plot. Twist: he's the protagonist

iirc, Claude Code has literal flags to detect frustration from the leak a few months ago, and I've since really stopped cursing at the LLM.

How about Microsoft fixes npm, github, and vscode

Woz rizz, Jobs dead

Good!

Y'all see the plan though, right? Control the internet. Remove e2e encryption. Control/stifle dissent and uprisings.

The plan is to replace the 99% with machines. It doesn't matter to the 1% if you survive their glorious, great filter.

The great filter is billionaires. It's a billionaire control problem, not a superintelligence control problem. You're livestock. There's a better ox and cart that just pulled down the gravel road.

nauseating to think that Amazon was only charged $2.5B when they made somewhere around $25-44B off the Iliad cancellation thing.

If the Mythos era isn't just hype marketing, "sec-scurity" might not be a valid strategy anymore. If you're taking a beat because you're small and irrelevant, you could still be massively fucked over a breach.

Mom & Pop code shops might be high risk if nation-state level vulnerability-exploitation becomes economically viable to any disgruntled prick.

Listen, I would actually be willing to support something like this, but Jesus Christ when will we put somebody in congress with a CS background who can literally just chime in and say "use Zero Knowledge Proofs for this, people might actually buy that you're not just building a surveillance state."

I wonder if this is happening because Mythos

A cybersecurity pandemic will surely be the Hiroshima that wakes people up to AI. /s

Agree. Maybe novel is the wrong term given my framing. "Listenable?" The combinatorial space of music is hyperastronomical, effectively infinite. And most of it is probably noise.

AI isn't "searching" in the standard indexing sense. But if say, Suno, is doing stable diffusion on fourier transform heat maps, and there's a finite space of configurations... it is using a heuristic approach to pick an option from a well-defined (gargantuan) set of options.

Exactly. Gorillaz was literally the cartoon characters in my mind for the entirety of my childhood, before I knew who Damon Albarn was.

I love music, I make music. It is sincere.

I'm not saying AI music is inherently good. I'm saying good music is good music.

My take is literalist, not reductive: Music (and any art product) is configuration. The media involved are the physical laws of nature that present the configuration.

It can be a means of communication, sharing human experience, ...

1. You're construing the journey with the result. The act of creating is not typically the result product itself.

2. Any perceived relationship to the artist via product is virtual and parasocial (respectfully stated). You have no relationship with Bach or Shakespeare or Michelangelo--you have an appreciation for their accumulated works. You may have a fascination for their stories and life through literary works. You have no relationship with your favorite artist, you have their works in your media library.

I see music as "the space of all possible 5-second clips at stereo 48kHz 24bit depth". If you think about it, that space contains the intro to Stairway to Heaven and Oops I Did It Again, and the end of either song. It contains every 5-second segment of O Fortuna, plus a previously unimagined O Fortuna Remix with MF Doom rapping the pledge of alliegance backwards. My point is, AI is getting good at searching music space for novel patterns, and that's entirely the point of music, not making a career out of being an alcoholic minstrel with a tour bus.

The audience will out the good patterns, and it's up to the musicians or AI companies to serve better patterns.

Used Suno to reimagine a handful of my old demos late last year, and honestly the results floored me. I could never release those tracks though, purely out of shame. But it seems pretty practical to study the AI remixes to understand what I like about them, and use these as a practice tool for music production.

I... wow, you made an LLM that can actually tell jokes?

This shark story the book opens with seems... implausibly Hollywood. But alright I guess.