This repository has been migrated to the self-hosted ari-web Forgejo instance: https://git.ari.lt/ari/enterprise-add

Enterprise add

Have you noticed how companies shove AI into EVERYTHING? I find it very annoying. But why not roleplay as them?

This program, written in pure C99 uses machine learning to predict a + b because why the fuck not! Maybe companies will adopt this enterprise add and incorporate it into their products because it's ✨ AI ✨

Watch how this two-neuron single-layer model was developed in 30 minutes on YouTube (live time): https://www.youtube.com/watch?v=6md1n5KaHMo

Dependencies

• A C compiler
• GNU make
• OpenSSL
• A computer which can run thousands of passes of gradient descent and some extra

This could have absolutely no dependencies, but it's enterprise. You wouldn't get it.

Licensing

Enterprise add is licensed under GPL-3.0-only by Ari Archer <ari@ari.lt>

Running

For best 🚀 performance 🚀 run it like this:

export STRIP=llvm-strip
export STRIPFLAGS='--strip-debug --strip-sections --strip-unneeded -T --remove-section=.note.gnu.gold-version --remove-section=.note --strip-all --discard-locals --remove-section=.gnu.version --remove-section=.eh_frame --remove-section=.note.gnu.build-id --remove-section=.note.ABI-tag --strip-symbol=__gmon_start__ --strip-all-gnu --remove-section=.comment --remove-section=.eh_frame_ptr --discard-all'
export CC=clang
export CFLAGS='-Wpedantic -flto=full -fno-trapping-math -fstrict-aliasing -fno-math-errno -fno-stack-check -fno-strict-overflow -funroll-loops -fno-stack-protector -fvisibility-inlines-hidden -mfancy-math-387 -fomit-frame-pointer -fstrict-overflow -Wshadow -fno-exceptions -D_FORTIFY_SOURCE=0 -Wall -Wextra -fno-signed-zeros -fno-strict-aliasing -pedantic -Ofast -fvisibility=hidden -ffast-math -funsafe-math-optimizations -std=c99 -fno-asynchronous-unwind-tables -Werror -fdiscard-value-names -femit-all-decls -fmerge-all-constants -fno-use-cxa-atexit -fno-use-init-array -march=native -mtune=native -pedantic-errors'
make -j12
make strip

Performance

The performance is very fast. It can add 2 + 3 with amazing accuracy only after 300 thousand epochs and a very miniscule precision loss of 0.03 but who cares about that - it's AI ✨:

2.000000000000000 + 3.000000000000000 = ~4.977217190454309

It is a huge leap in the industry of adding two numbers together.