Google DeepMind Introduces Direct Reward Fine-Tuning (DRaFT): An Effective Artificial Intelligence Method for Fine-Tuning Diffusion Models to Maximize Differentiable Reward Functions