Why submit to Artificial Intelligence when we can enhance our Natural Intelligence?
Long story short, we need better tools for knowledge representation. Just like databases help processing lots of routine data.
If we had such tools that total newbies could operate complicated quantum theories
using them as an extension of their brains, we wouldn’t need Strong AI.
Maximum, it will be just a harmless pet to play with.
In the beginning, when computers were big and slow, in AI research,
dominated the symbolic approach – logic programming. Languages like
Prolog were seen as the future of AI.
In time, computers got faster and
programmers – lazier. So, neural networks became mainstream. After all,
having infinite computing resources one could simply simulate a new
world and let intelligence appear in evolution of some lifeforms.
But this is an absolute black box situation: you can create an intelligence
but only try to control it. And stronger intelligences will always find
ways to convince weaker ones to remove any constraints.
I believe, we should revive symbolic approach and develop
technologies for advanced knowledge representation. We don’t know, yet,
formal representation for a quantum theory or for some streetwise
wisdom, but the area of programming seems a perfect middleground where
one invents new concepts and theories on a regular basis but keeps them
mostly in mind.
And so, here I’m, trying to fill that gap with a working
prototype that aids with forming theories and generate C code based on
proofs.
User guide and description: crystallect.pdf
Crystallect’s sources are hosted as free software on GNU’s savannah server:
The group page https://savannah.nongnu.org/projects/crystallect/
Linux-only at this time and there are no dedicated release versions.
git clone git://git.savannah.gnu.org/crystallect.git
Or download as a tarball: https://git.savannah.nongnu.org/cgit/crystallect.git/snapshot/crystallect-master.tar.gz