Obviously there's no "evidence". Why would you even think we need AGI? But I'm happy to hear your reasoning if you were one of the few/only? people who imagined that software that could predict the next word could do what it now is doing.