> With each new release, they add evidence that they were correct.
Is it though? If the goal is human-level AI, or hell, even rat-level AI, the evidence is pretty convincing that you should be able to train and deploy it without requiring enough energy to sail a loaded container ship across the Pacific Ocean. Our brains draw about 20 watts, remember. This suggests to me that no, in fact, scale will not get us "there".
Is it though? If the goal is human-level AI, or hell, even rat-level AI, the evidence is pretty convincing that you should be able to train and deploy it without requiring enough energy to sail a loaded container ship across the Pacific Ocean. Our brains draw about 20 watts, remember. This suggests to me that no, in fact, scale will not get us "there".
https://www.forbes.com/sites/robtoews/2020/06/17/deep-learni...