>In fact, simple measures like forbidding deployments of models with more than a certain number of parameters would probably slow things down a lot. Are we eve

>In fact, simple measures like forbidding deployments of models with more than a certain number of parameters would probably slow things down a lot. Are we even sure GPT-4 has dramatically more params than GPT-3? They have stopped revealing the parameter counts with 3.5.