Part 4/8:
To shift from 20% automation to full 100% automation, Davidson’s model evaluates two types of resources: the need for additional computing power and the effectiveness of enhanced algorithms. The uncertainties surrounding the estimates hint at a range that could demand as low as ten times more resources or as high as 100 million times more.
The resource requirements are staggering, with some estimates suggesting that reaching the 100%-AI threshold by 2043 could demand up to (10^{36}) FLOPs of computation, the equivalent of a training run that would need to be initiated back in the Jurassic period for completion today.