[rvz]

Deep Learning (DL) is just as great at wasting tons of electricity and gallons of water [0]. Crypto at least has found an alternative and less wasteful methods than proof-of-work that is viable for them to migrate and switch to.

Unlike the field of Deep Learning, which after decades still has not found any viable alternatives to their methods of training, fine-tuning and inferencing operations and continues to waste insurmountable amounts of resources.

The worse part is the outputs after training from such black box AI models; including hallucinations, lack of transparent explainability and a 'high accuracy' of being confidently wrong.

There is nothing 'useful' for a hallucinating glorified unchecked bullshit generator, or a vision system getting confused over a single pixel.

[0] https://gizmodo.com/chatgpt-ai-water-185000-gallons-training...