Today's most advanced AI systems require around 25,000 GPUs requiring about 10Megawatts of power. Each GPU has about 80 billion transistors. So we talking about 2^14 transistors, and each one can cycle on off more than a billion times in a second.

By contrast, the human brain has about 66 billion brain cells and uses about 30 watts of power.

Seems to me the AI algorithm must be tremendously inefficient to require this much hardware and energy to only do part of what a human can do even though it can do what it can do much faster than a human.