Contextual reasoning is very important for many situations, because time is one of the most important contexts. You can have a fact like "the iPhone is the best selling phone". This fact is true or false based on the context. It is definitely false in the real world year 2006 and before, because the first iPhone was not released yet. While the fact may be true right now in a global context, if we are talking about a specific country or region, it may be false. For example, I would expect that satellite-capable phones may be more popular in a highly remote (non-urban) country or region.
I've been refreshing myself on ontology design and knowledge representation lately. I'm kind of surprised that is hasn't moved very much in the last two decades. I'm guessing we're about to hit an inflection point there as it becomes a cool topic with the current AI trends.
It occurs to me that if we want LLMs to be able to reason better, we may need to write texts that explicitly embody the reasoning we desire, and not just use a bunch of books or whatever scraped from the Internet.
the tutorial page https://www.idp-z3.be/tutorial.html is a bit more enlightening
as is https://interactive-idp.gitlab.io/ learning material
but I didn't find those first, I found the docs first (via the four "learn more" links on the landing page)
and I don't think they are the best starting point
https://docs.idp-z3.be/en/latest/introduction.html
> IDP-Z3 can be installed using the python package ecosystem.
> install python 3, with pip3, making sure that python3 is in the PATH.
> use git to clone https://gitlab.com/krr/IDP-Z3 to a directory on your machine
> (For Linux and MacOS) open a terminal in that directory and run the following command
First thought is... why isn't it published to PyPI?
But then elsewhere https://docs.idp-z3.be/en/latest/IDP-Z3.html
> the idp_engine package available on Pypi.
And looking in the git repo it seems like this is the same package published as https://pypi.org/project/idp-engine/
Why not just say "pip install idp-engine"...?
Quick response: with pypi, you can only install the reasoning engine. By cloning the repository, you get the full suite of tools, including the "interactive consultant".
(I have nothing to do with IDP-Z3) This is the general pattern:
* If you are a developer: get sources and work on sources, because, probably, you want to add to them or modify them etc. Part of this experience may be creating a package and installing it with pip or some other package installation program, but it's not usually what other developers (contributors) would be working on.
* If you are a user: use the supported package manager to obtain the package and use it through documented interface.
I'm honestly not sure if this is why or how these instructions came to be, but this is a fairly common thing to do, so, w/o looking, I'd guess this is what it is.
For example, you give IDP-Z3 the formula that links a tax-free amount, a tax rate and a tax-included amount, and the values of any two of its parameters, and it will compute the missing parameter. You do not need to write 3 different formula, one for each case. If you give him only one parameter, it will say that the other two parameters are relevant.
