> I’ve read many articles on the same topic but never found it presented this way - hopefully N + 1 is of help to someone.
Can confirm; yes it was helpful! I've never thought seriously about parsing and I've read occasionally (casually) about Pratt parsing, but this is the first time it seemed like an intuitive idea I'll remember.
(Then I confused myself by following some references and remembering the term "precedence climbing" and reading e.g. https://www.engr.mun.ca/~theo/Misc/pratt_parsing.htm by the person who coined that term, but nevermind — the original post here has still given me an idea I think I'll remember.)
> This is an unabridged version of a solicited article for a forthcoming Blackwell Companion to the Philosophy of Mathematics. […] took over a year to write – which means, at the current pace of development in the field, that some of it is already slightly out of date.
> There is always a significant chance all of it is leaked sooner or later.
As an adversarial/worst-case model, it can be useful to think of every service as potentially storing forever all the data that you ever give it access to. As a practical matter, services have terms of service that they follow. If your Claude Code terms say that your data will not be used for training, you can be reasonably confident that they will not be, and storing the raw inputs forever (as suggested by “significant chance all of it is leaked sooner or later”) would be even more unlikely. (For example, Google has entire teams dedicated to compliance with users' “wipeout” settings. You can take a look at https://myactivity.google.com and https://myadcenter.google.com to see some of what Google knows and thinks about you, and if you've chosen "Auto-Delete after 3 months" or whatever, you can be very sure it will be gone after that time. Every single team that stores user data is required to comply with this.)
I do think the services make it harder than it should be, to find out what the terms are — for a given usage of their services whether and for how long the details will be stored by them. Just saying that you can find this out and generally rely on it at least at the time (at a reasonable threat model, e.g. not treating the service as a malicious adversary having a giant law-breaking conspiracy that has never been exposed).
To me this is the money shot (but it takes a couple of passes to understand):
> No small amount of criticism of LLMs is downstream of past decisions to reify form over function, resulting in the substance having been optimized out. Now the LLM threatens to make the form available in seconds
Whether or not one can tell it's AI generated, one can certainly tell it's not Knuth. For one thing, the writing style is very different. Not that there haven't been other great computer scientists who may have written in this style, but it definitely doesn't sound like Knuth (there is no "being a bit cheeky" for sure). But also, the ideas it has produced are simply more of the same; kind of a natural progression / what a typical grad student may write. Knuth always has something new and surprising to say in every paragraph, he wouldn't harp on a theme like this. Also he mixes “levels” between very high and very low, while the paragraphs you quoted stay at a uniform level.
But of course, writing as good as a grad student's (just not the particular delightful idiosyncratic style of a specific person) is still very impressive and amazing, so your concerns are still valid.
I think it's also a privacy thing; you have to go into the Waymo app and “connect” your YouTube Music account (even though both have the same @gmail.com address), because otherwise the terms of service of one do not allow sharing data with the other without user consent. (Contrary to popular perception Google is very finicky about privacy, at least privacy as defined as conforming to the terms of service.)
> Falling doesn't have to be dangerous. You can fall a lot without getting hurt, if you learn to fall safely. With inline skating, you have protective gear (helmet, knee/elbow pads, wrist guards) which protect you, and you have techniques for falling which let you use this gear to its fullest potential.
Is that actually true? Is it possible with enough protective gear, that falling can be safe, even for older people? Doesn't your own body weight come into the picture, despite helmets and knee pads? (Genuinely curious!)
IME yes, it absolutely can be. I am approaching middle age & still comfortably enjoy pushing myself in physical activities where falls are likely, with zero significant injuries aside from a couple sprained ankles from playing basketball (& technically the ankle rolling came before the fall in these couple mishaps; letting my body roll/fall out of it just helped reduce the severity). Also it's more about technique & familiarity/reflex training than safety gear, although I do wear a Zamst ankle brace on my weak ankle whenever I play basketball & started wearing a helmet for snowboarding a few years ago. Jackie Chan & Buster Keaton were even better at this, although they pushed it a lot farther & did sustain major injuries in their stunt careers.
However, there's a big caveat: I've been practicing falling safely since a young age & really mastered it in my teenage years practicing martial arts & snowboarding. I'm sure it's much harder & more dangerous to learn if you first start in middle age, although I'd imagine it's still possible with the right training & appropriate caution.
Thank you. Yeah given the caveat I think it's probably hard then, unfortunately. (For context, I'm someone who's generally very uncoordinated, didn't play any sports growing up, etc, and a few months ago at 39 I fell from kitchen-counter height or possibly even just footstool-height and somehow managed to fall awkwardly on my side and fracture my hip (acetabulum), which took a couple of months to heal. I'm told that this kind of fracture is unlikely in people this age unless there's high-speed impact or osteoporosis involved, but well, I have a talent for awkwardness.)
The broader point of the post I actually agree with though, but the lesson I'd take away is to engineer environments such that it's ok to fall/fail safely.
reply