I've always wondered if it would work out well for an LLM to evaluate a constraint program/script when it encountered a problem involving constraints / logic. like or-tools with python or evaluating a minizinc program.
Yes, that's very plausible and I've thought similarly but not exactly the same approach for the diagramming limitations of LLM [1].
Fun facts, the presenter John Hooker was asked about how to determine the suitability of a particular heuristic solver for specific problems in his presentation. He casually answered that if he knows the solutions to that he will probably win a Nobel Prize. But perhaps AI/LLM can help in a way to recommend the solver based on the type of applications or problems.
If I'm not mistaken there's also Donald Knuth (TAOCP) asking questions after the JH's presentation, how often you see that?
I also don't prefer using strings, but to be fair, HTTP methods are just strings when the request is received. There is some beauty in that in matches the prefix of the first line of an HTTP packet
Since you seem interested, may I spur you in a perhaps-not-thought-of direction?
I’m currently working on a hand-wired board. Since you mention pcb development, I don’t feel that hand-wiring is too much work to mention. The soldering and electronics work is trivial (nothing small). I was attracted because it allows me to move the customization into the 3rd dimension with a Dactyl-manuform clone (I’ve actually hacked some closure together).
I’ll make sure to post where appropriate once I’m finished. I’m basically hand-wiring a corne (CRKBD, so there are guides available) only I’ve created a custom 3D manifold fit to my hand.
This is cool! I've had similar ideas, but when it comes to creating the curriculum it was always the time consuming part. If the graph was in something like a wiki, where people could contribute, add linkages, add new lessons, etc., and a system to automatically prune the graph to each student, there could be a globally sourced and maintained "school" for all human knowledge.
The courses are markdown and JSON files in GitHub so close enough. It does restrict a bit because it assumes some technical proficiency, but there's probably some middle ground (open a GitHub issue to suggest changes and have someone else write the PR, for example).
I find the lack of learning science and the state of teaching how to learning to be detrimental to our current society where people are required to know more and more. Knowledge that doctors, lawyers, even computer engineers keep increasing as time goes on, but the way we learn has never been scrutinized or emphasized and is mostly up to each person to deal with.
I mostly agree, learning efficiency is not directly tied to the method, but how the brain processes it over time and may even require multiple methods to sufficiently learn something. I wouldn't be surprised if emotions or feeling frustrated while trying to learn hampers it as well.
I think by design they're not meant to be associated with a single user. A hospital would publish all id's of users that have found positive with covid and each other individual user's themselves will see if they have been in contact recently with one since they have saved all id's they've seen recently
Could network analysis be used to deanonymize these rolling ids? I remember reading that network analysis easily deanonymizes blockchains. Seems same would apply to rolling ids.
JSON has no notion of object "type". So if you have a list of things, whose processing is determined by their type, you may need to parse the entire object before finding the key/value pair that distinguishes its type.
Because XML puts the "type" of an object syntactically first, it supports this use case.
Pity they don't use it for npm, which cannot create a local search index because of this: it tries to load a large JSON file into memory and fails. I never was able to do this on my VM despite trying to give it more memory (up to 2 GB, at which point I gave up).
