Everyone points to the convenience, which is compelling. But Dr. Marshall's post is about how the AI is deployed, which is at the heart of many of the legal disputes about copyright and ownership. If I am asking for it to create an image of a Toyota Supra (which i did in GPT-4 Plus), I have no idea where it derived the image from. That aspect becomes an issue for copyrighted works - it's why many people are recommending to not upload proprietary code into Bard, ChatGPT. It is a good pair programmer, but it does force consideration of what should be used on the platforms.