It's about time

No.14022038 ViewReplyOriginalReport
Ok listen up you high IQ ding dongs. This """AI""" group cuck jerkoff session needs to stop right now. And I personally will no longer tolerate people telling me about their "AI Solution" or their "machine learning algorithm" without them also showing me the resulting intelligence explosion from their algorithm that AUTOMATED INTELLIGENCE. If you want to be a machine learning engineer, be a machine learning engineer and dont cower in the corners of pytorch documentation wetting your pants and training your stupid little models watching the epoch counter go up like some kneel de-grass tyson npc

If no one else is gonna say it then I'm gonna say it: AI needs to be developed in a simulation. PERIOD. Aint no fuckin way you're shoving in matrices on one end and getting "INTELLIGENCE" on the other side.
We need a recursive algorithm that operates over time so it can interact with other beings IN TIME. -|LIKE|- -|US|- . And offer useful solutions to problems in plain english.

I'm gonna spell it out nice and slow for any retards still reading:
1) Use a spiking neural network framework for the underlying simulation.
2) Program a simple simulation like a top down minecraft or something where the player can clearly make changes to the world, change its perspective, and IMPLEMENT solutions that would be useful to its goal which would all be relative to how the punishment and reward signals are set up.
3) In a genetic algorithm define genes consisting of 30 (or about 30 idk find out for yourself you lazy dipshit) or so numbers that dictate how adjacent cells in the SNN framework interact with each other,
4) Set up the internal agent's body needs like getting food blocks, avoiding enemy blocks that damage you and compile them into interpretable signals to the SNN, do the same with the nearest 'n' blocks around the entity, then make the output control what the entity ends up doing in the simulation (move or place a block)

continuning: