Compression, creatively reconceiving problems, RL, interpretability, model architecture, model modification, democratization of AI, machines of loving grace type scenarios.