Is There a Benefit to Having a Thousand GPT3s?

In the next bullet point I got here, the poor B estimates that in their expectation, each hyperscaler will have somewhere in the range of 10,000 to 200,000 GPT3s equivalent within five years. And he says to be clear, those parameters is not functionally equivalent to what synapse, and this comparison is not an attempt to conclude we will have human level intelligence. Nor am I suggesting that scaling a parameter count in a transformer is the correct use of that compute budget. But he's pointing out that this is a really, really big number,. And five years is not that long a time. We have both at the same time. Even if we didn't have