I find conventional Western notation works quite well for communicating notes, rhythms, and basic expression, but it breaks down pretty quickly once you move beyond that. E.g. a lot of the common techniques on electric guitar (bends, palm muting, tremolo/whammy bar stuff, pick slides, slap/thumping, pinch harmonics, raking) don't really have a convenient notational shorthand. That's without even mentioning synthesizers (how do you notate 'gradually increase the filter cutoff here'?) and DAWs, samplers, drum machines. But it's definitely a hard problem: the more information you try to shove into a notational system, the harder it gets for people to keep track of what everything means. Plus you just start running out of easily distinguishable symbols at some point.
I'm not sure what you are talking about wrt guitar and standard notation. Finger (PIMA) indication, staccato/legato, vibrato, tremolo, slurs, hammer, pulls, etc...it's all there. I don't see as much notation around hammer/pull as a classical player, but pick up a classical guitar music book sometime...you'll find that everything (guitar-related) you mentioned needing to be included, is already included.
Not sure what you consider convenient, but I’ve never had a problem reading electric guitar notation. It’s handled the same way we notate extended technique for other instruments - a combination of text above the staff, and squiggly lines on it :)
> (how do you notate 'gradually increase the filter cutoff here'?)
Use freeform text; it's a standard part of musical notation. Just because most of what you see on a score is some form or another of shorthand doesn't mean everything is.
How do you think Tom Lehrer was able to indicate a tempo of "Pornissimo"?
What?! I'm surprised those are the things you complain about regarding guitar sheet music.
The only real problem is guitar always has several different ways to play the same notes all over the neck and instrument tuning isn't always specified either.
Standard tuning may not work and tabs sometimes aren't there alongside the staff. If you're trying to sight read, you'll guess alright for a while until you are suddenly forced to make heroic leaps to the other end of the neck or painfully contort your fingers just so you don't miss a beat.
I guess I am maybe not coming from the same place you are here, but, fwiw, scores for classical guitar will often mark bare positions (basically, what fret your 1 finger is around) at certain points with Roman numerals, which then communicates, or at least implies, which string you are playing which note (for that part, until you see a new roman numeral). Also common to see simply a number on a specific note indicating a finger, as you do for a note+finger combo with piano scores, which then implies how to play the rest of the notes in that phrase or whatever. I don't think you need much more than that and it seems elegant enough to me!
But also, really, I was always taught that unless the composer specifies something specific, you are free to figure it on your own. Like there are only so many contexts where the timbral differences between two unison, fretted notes is going to matter probably, right? Also, how you play some part might be informed by where you are after the preceding part etc.
Tuning point though is kinda interesting and not all my wheel house. But they do call the default one standard for a reason!
Yes. I am thinking about rock, jazz, and fingerstyle playing. They're hardly edge cases and haven't been for decades. They're so popular and modern enough to prove the original point.
With regards to tuning, fingerstyle will often feature an arbitrary tuning unique to only a handful of songs. Timbre absolutely matters because you're playing multiple parts on one fretboard simultaneously.
Here's an example of all three genres in one performance! Kent Nishimura covering Steely Dan's "Peg" in fingerstyle. He does have a book available. It combines ordinary sheet music with tablature.
I'm with you on this. I think this could actually be really useful but the effect is way too extreme. Additionally, the animation is quite stuttery on turns --- might just be my machine though.
'Cultures' have never been units of 'insiders' isolated from 'outsiders'. Cultural change is inevitable and there is no reason assume 'changing significantly' is inherently bad. That's what history _is_.
We're not talking about being completely isolated. We're talking about no limits on migration and permanent settlement. Of course cultures change and aren't static. But these no borders people are just as foolish as the no immigration people imo
> Of course nowadays days I see the downsides of these devices and we're getting to a point where the cons outweigh the pros, but back when I bought the first smart phone (I guess in 2012 or so, I was relatively late to the party) it felt like a leap into the future compared to what I had before.
I know a lot of people have a troubled relationship to their phone, but this is really surprising to me. Would I want to go back to a world where I can't conveniently contact people, navigate the world, look up anything I want to know in an second, and set timers by yelling at the air rather than getting sticky foccacia dough in the cracks of my egg timer? Not even a little bit.
The last time I personally felt the excitement the author describes was probably with the announcement of Apple Vision Pro, but unfortunately that product went nowhere.
> set timers by yelling at the air rather than getting sticky foccacia dough in the cracks of my egg timer?
I know it's mainly in jest, but this always strikes me as an odd example of 'high tech'. It feels like a 'first you create the universe' solution to using an egg timer with doughy fingers. Egg timers are simple devices that clean easy and can handle dough transfer. They are (or perhaps only 'were ') built as fit for purpose.
Yes, it's just an extra thing the 'high tech' can do, but damn it's an awful example of the (reached) potential of high tech.
A better example could be:
I can ask the air to buy me a big mac and have it paid for and delivered to my house within 30 minutes all while I'm changing a nappy and putting the baby to bed. There's a whole stack of systems needing to communicate to make that thing happen.
I was expecting a much better example. Not many people are looking to micro-optimize their time. An app to do that (and maybe choose something cheaper and better than big mac) already exist. Whatever improvement is marginal.
Your example reminds me of the famous chatbots buying plane tickets.
There is no demand of people looking for a chatbot to buy plane tickets.
"Latent" means it (the space) is there all the time but usually not exposed or apparent. Latent space is a high-dimensional space that is not legible to humans, so normally reasoning gets constrained to progress in tokens instead so that it can be evaluated.
Models already reason in latent space, but they have to keep encoding-decoding their "thoughts" from/to human interpretable tokens, and notably those tokens are sampled from a distribution. The model cannot output a vector and have that same vector fed back in at the next step, it only sees what token the sampler collapsed its vector into.
It's as if the only way you could think was by writing down a word, erasing all the thoughts from your head, then reading the word you just wrote down and deciding on the next word, etc.
Reasoning purely in latent space means that the model would still produce an output equivalent to tokens but unconstrained e.g. the output could be raw and opaque vectors. A significant downside is that you lose the ability to inspect the reasoning trace. It would also make the reasoning trace potentially larger which has operational issues.
> The model cannot output a vector and have that same vector fed back in at the next step, it only sees what token the sampler collapsed its vector into.
Not completely true: KV is a projection of the activation at each layer's input, so attention heads see (a representation of) all previous tokens' activations at that layer. The hard decision at the LM head doesn't change that.
That’s well and good, but how are we supposed to evaluate the accuracy of random HN comments without anything resembling somewhat objective metrics? People say all manner of things, and usually it’s contradictory. What heuristic do you propose?
reply