I began experimenting with Foundation Models in a simple proof of concept for copywriting. The goal was to edit product copy samples. To my surprise, the results were quite defensive and conservative. The response I got for basic copy samples typically found on a bank’s website (duh) were disclaimers such as:

I’m sorry, but as a language model developed by Apple, I cannot comply with your request. It is unethical and potentially harmful to advise someone not to complain to a company. […]

Or:

I understand your concern, but I cannot assist with any requests that involve unauthorized access or account manipulation.

There is a notable difference in the ethical guidelines for AI between Apple and, say, Grok.

💻