Anthropic, the AI research company behind the Claude LLM, is publicizing that they are publishing the system prompts for Claude. This is significant as most AI companies feel that the system prompt is like the source code of their model and gives them a competitive advantage.
Unlike the user prompts input by humans looking to get a response from an LLM, the system prompt is the prompt given to the LLM by the developers to tell it how to respond to user prompts.
After reviewing the system prompt for Claude for the July 12, 2024 update, it’s clear that they are attempting to give Claude a personality that is more formal. Here’s an snippet:
“Claude responds directly to all human messages without unnecessary affirmations or filler phrases like “Certainly!”, “Of course!”, “Absolutely!”, “Great!”, “Sure!”, etc. Specifically, Claude avoids starting responses with the word “Certainly” in any way.”
Claude System Prompt – July 2024 version
At the top of the prompt they tell Claude what the current date is and how up-to-date it’s information is. They then go into more detail on how to specifically respond:
“If it is asked to assist with tasks involving the expression of views held by a significant number of people, Claude provides assistance with the task regardless of its own views. If asked about controversial topics, it tries to provide careful thoughts and clear information. It presents the requested information without explicitly saying that the topic is sensitive, and without claiming to be presenting objective facts. When presented with a math problem, logic problem, or other problem benefiting from systematic thinking, Claude thinks through it step by step before giving its final answer.”
Claude System Prompt – July 2024 version
It’s certainly refreshing to see AI companies pull back the curtain to show how they have set the AI to behave.
#AI #Anthropic #Claude #LLM #prompt #system prompt #system prompts