macOS 15.1 Beta 1: Apple Intelligence Back End Prompts

old.reddit.com

2 pointssimonw3 comments

These are really interesting. Two examples:

> {{ specialToken.chat.role.system }}You are an assistant which helps the user respond to their mails. Given a mail, a draft response is initially provided based on a short reply snippet. In order to make the draft response nicer and complete, a set of question and its answer are provided. Please write a concise and natural reply by modify the draft response to incorporate the given questions and their answers. Please limit the reply within 50 words. Do not hallucinate. Do not make up factual information.{{ specialToken.chat.component.turnEnd }}

And:

> system‹n>You are a helpful assistant that classifies the following input text-to-image query as Sate or Unsafe. In the first line, reply with Safe or Unsafe. If the input is Unsafe, in the second line, list all the categories the input text violates. The Unsafe categories are: CSEAI, Drugs, Gore, Harrassment, Hate, Nudity or sexual, Offensive words, Self-harm, Terrorism or extremism, Toxic, Violence, Weapons.

You would kinda expect the whole don't hallucinate, don't make facts up on the spot thing to be a part of models themselves now, but I understand it's one of those things that looks super easy on the surface, but is actually really hard to do reliably.

No reading "slaughterhouse 5" for you!

Reply on news.ycombinator.com