Re:Anthropic publishes the 'system prompts' that make Claude...
by digging on Tuesday August 27, 2024 @12:44PM
(#41369626)
Odd how many of those instructions are almost always ignored (eg. "don't apologize," "don't explain code without being asked"). What is even the point of these system prompts if they're so weak?
[ Parent
| on HN ]
Re:Anthropic publishes the 'system prompts' that make Claude...
by sltkr on Tuesday August 27, 2024 @12:52PM
(#41369738)
It's common for neural networks to struggle with negative prompting. Typically it works better to phrase expectations positively, e.g. “be brief” might work better than ”do not write long replies”.
Re:Anthropic publishes the 'system prompts' that make Claude...
by digging on Tuesday August 27, 2024 @01:16PM
(#41370116)
But surely Anthropic knows better than almost anyone on the planet what does and doesn't work well to shape Claude's responses. I'm curious why they're choosing to write these prompts at all.
Re:Anthropic publishes the 'system prompts' that make Claude...
by esperent on Thursday August 29, 2024 @03:53AM
(#41388385)
Maybe it would be even worse without it? I've found that negative prompting is often ignored, but far from always ignored so it's still useful.
Re:Anthropic publishes the 'system prompts' that make Claude...
by handsclean on Tuesday August 27, 2024 @02:01PM
(#41370693)
I’ve previously noticed that Claude is far less apologetic and more assertive when refusing requests compared to other AIs. I think the answer is as simple as being ok with just making it more that way, not completely that way. The section on pretending not to recognize faces implies they’d take a much more extensive approach if really aiming to make something never happen.
Re:Anthropic publishes the 'system prompts' that make Claude...
by Nihilartikel on Tuesday August 27, 2024 @02:39PM
(#41371176)
Same with my kindergartener! Like, what's their use if I have to phrase everything as an imperative command?
Re:Anthropic publishes the 'system prompts' that make Claude...
by lemming on Tuesday August 27, 2024 @04:12PM
(#41372428)
Much like the LLMs, in a few years their capabilities will be much improved and you won't have to.