h a l f b a k e r ycarpe demi
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Intermedia Equivilincia
Compilers translate one specification into another thing. Every person has different skills. There more than one way of doing them or thinking about same thing. Let's translate between them. And detect equivalencies. | |
Problem is, everyone understands different things intuitively differently.
I ask you to go read the Linux kernel or Postgres codebase.
Can you understand it? Or is it too big for you to understand?
The number of people who can casually read the linux codebase or postgres codebase is small I
think.
Ask someone to solve a logistics problem, graph problem, date problem (calculate how many times the bus route happens), or social problem, and you'll get different understandings of the problem.
Intermedia equivalincia is a database of mental representations of things.
Imagine an intermedia equivalincia of mathematics, different ways of doing mental arithmetic numbers and thinking about things.
There should be an app that lets me pick a problem and see a representation of it and then directly interact with that model and see equivalents of that.
It's compiler intermediate intermediate representation. You can translate between languages.
Open AI Sora
https://www.youtube...?v=h8Bi1O3nTDU&t=1s
Callum Upton, YouTube [Skewed, Mar 11 2024]
ChatGPT fake lawyer
https://www.youtube...watch?v=sEOapG7-kro
Steve Kehto, Youtube [Skewed, Mar 11 2024]
AI generated games
https://www.youtube...watch?v=jpIRMS-c-dU
Callum Upton, YouTube [Skewed, Mar 11 2024]
[link]
|
| |
I think this is basically a polygraphy, a kind of mathematical or logical intermediary between different natural languages. Umberto Eco deals with these in his 1997 book 'The Search for the Perfect Language'; on p196 he disusses Kircher's 'Polygraphia' of 1663. The conclusion is that it is a stupidly impractical idea, that cannot work for many different reasons. |
|
| |
That's not this idea, this idea is that I can draw a representation of something that I understand and share that, and it maps back to the original semantics. |
|
| |
This is what compilers do. |
|
| |
We have people who intuitively understand what they want to do and people who understand details. |
|
| |
Currently all the work is one way. Programmers encode movements between things to implement the first person's requirements. |
|
| |
Instead, we create easy to understand models, such as tables lists or point and click graph editors and let people encode their models in other people's intermedia. |
|
| |
I parallelize expensive tasks with graphs for example, because whenever a graph splits, you can do any number of edges in parallel. |
|
| |
This is is that everyone can produce quizes or questionnaires for others asking them to encode what they want in a representation because that representation is useful to other people. It's "equivalent" |
|
| |
I would give people a questionnaire for pages they want on their website, or what the order of operations should be . Or a list of events. |
|
| |
Right but Eco's point is that universal semantic mapping is not possible. You would enjoy his book I think, it is an overview of a millennium or two of people trying to do the kind of things you are interested in, with detailed description of their methods and principles and a succinct technical analysis of why they failed. |
|
| |
//a stupidly impractical idea, that cannot work for many different reasons// [+] |
|
| |
A related idea, that *could* work, is not a single language that captures everything, but an evolving collection of models that captures subsets of different languages, where the subsets are bounded by task-related contexts. |
|
| |
This idea would be really helpful for my daily work. |
|
| |
Someone captures their idea of how something works as a graph or a table or a list, |
|
| |
then I translate that into something else and the translation is automatically stored and processed. |
|
| |
Currently, people spend a lot of time interpreting words and then doing things that are unrepeatable. |
|
| |
It's basically algebra, has anyone heard of term rewriting systems? |
|
| |
Complex translations or movements between terms. |
|
| |
[pertinax] are you not simply describing natural language hybridisation, pidgins, and technical vocabularies there? |
|
| |
[chron] algebra is widely-known-to-exist, and is already used where it is the best way to express ideas and specifications. For example I will order 500 small metal components to be custom made for me; I could happily write a detailed 2000-word description that totally specifies the object I need, but it is easier for me and for the engineering company if I send them a technical drawing marked up with numbers and symbols for the various parts of the machining. |
|
| |
I note that this idea is written in natural language rather than any such symbolic or algebraic system. |
|
| |
There is something really special about natural languages, in that they evolve over millenia in human brains, and are perfectly tuned to allow human brains to communicate with each other. Synthetic or artificial languages are basically useless "cargo cult" simulacra of natural languages. |
|
| |
Eco explains all this in lurid detail. |
|
| |
// interpreting words and then doing things that are unrepeatable // |
|
| |
That sounds rather romantic. Or, at least, Heracleitean. |
|
| |
//hybridisation, pidgins, and technical vocabularies// |
|
| |
Yes, but these things could be done better. |
|
| |
//term rewriting systems// |
|
| |
These take the hierarchy of terms as given. For your use cases, that's begging the question; in systems integration, one of the fastest ways to make a common object model that's broken beyond repair is to assume that the end- point systems use equivalent hierarchies. |
|
| |
//Eco explains all this// |
|
| |
I have not yet read 'The Search for the Perfect Language', and will add it to my to-do list. However, I am inclined to be suspicious of Eco's conclusions; like several other intellectuals of his generation (and the adjacent generations), Eco seems to have had a bit of a prejudice against any power structure larger than a clique. Thence, the prejudice extends to any semantic structure which might transcend cliques. |
|
| |
For example, there's a passage in Eco's "Name of the Rose" where he riffs on the incompatibility of homonymous units of measure across late-medieval Europe. We read this and are very impressed at first by this learned exposition of the contingency of everything and of the absence of any stable "out there" objective reality. But then we remember that, since that time, units of measure have been harmonized to a very high degree; we all* use the metric system, and, although the standard metre is not as romantic as the lesser Etruscan cloth-yard**, it has great practical utility, and its creation was not at all futile. |
|
| |
**No, that's not a real thing, but I'm too lazy to look up Eco's actual examples. |
|
| |
//A related idea .. an evolving collection of models that captures subsets of different languages, where the subsets are bounded by task-related contexts// |
|
| |
So a bit of software that writes and codes a game or other software for you in the programming language of your choice based on verbal input in spoken English (or any other language if the new and improved Google translate is plugged in), yeah, it's sort of where (what we laughingly call) AI is headed, it may not be what anyone could call intelligent without bastardizing the meaning of the word to mean something else entirely but what the thing we have that we call AI is good at is 'learning' that sort of thing, I thought they were already well on the way to developing things like that. |
|
| |
Have you seen some of the newer "AI" art generators? |
|
| |
I haven't - would you like to recommend one? |
|
| |
Sora video app? creates video from text, I've popped a link to a Youtube from a guy called Callum where he looks at a clip produced by it, search the first 3 words and you'll find links to other videos etc. |
|
| |
For text you have untold numbers of AI story generators out there, just do a search for that. |
|
| |
ChatGPT can be used to do a whole lot more than chat, though some people are getting a bit ahead of the curve of it's current competence on that one, see link above about a lawyer who got in trouble with that ;) |
|
| |
ChatGPT is even beginning to get into coding (another Youtube from Callum on that in links above). |
|
| |