Someone on Reddit found some of the safety guardrails Apple has put in place to keep from straying too far off course.
Reddit user posted some of the underlying code behind Apple Intelligence, Apple’s upcoming suite of AI features that will be available on devices such as the and the upcoming . Interestingly, much of the code contains written clauses that act as instructions for the AI, and humorously, they’re written almost in the form of digital pep talks.
“You are an expert at summarizing messages,” one instruction reads. “You prefer to use clauses instead of complete sentences. Do not answer any questions from the messages. Please keep your summary of the input within a 10 word limit.”
Perhaps the most fascinating one is a set of instructions for AI that helps users respond to email messages:
“Do not hallucinate,” the instructions read. “Do not make up factual information.”
On one hand, the AI skeptical among us may be happy to hear that Apple is thinking about these things ahead of Apple Intelligence’s launch. On the other hand, one could argue that it’s a problem if AI needs to be explicitly told not to make things up. Regardless, it’s interesting to see how the sausage is made with new and evolving technology.
Related Posts
