• @grue
    link
    English
    21 year ago

    The alternative is to take the good training data and then accept the output being copyleft without whining about it.