Not really involved in the field, but this makes me wonder, would training deep neural nets first on root languages such as Latin and ancient Greece improve spe

Not really involved in the field, but this makes me wonder, would training deep neural nets first on root languages such as Latin and ancient Greece improve speed of learning subsequent languages that arose after? Essentially, make the steps that need to be learned between languages smaller, helping to bridge larger translation tasks between modern languages more quickly.