Worth noting that just because the system prompt says "based on the GPT-4 architecture" doesn't mean that the model is actually based on GPT-4! The goal of a system prompt is to influence the model to behave in certain ways, not to give it truthful information about itself
ETHICS
-
Corporate Self-Regulation Fails Safety Standards in Tech
By
–
If you leave it to companies to decide what is safe you get the Boeing 737 max.
-
Should We Trust What AI Models Say About Themselves?
By
–
I tend not to believe anything a model tells me about itself!
-
Extracting AI System Prompts: Challenges and Hallucination Risks
By
–
I'm not sure that's the real system prompt – looks to me like it could be a hallucination based on training data. My own attempts to extract the system prompt have so far failed:
-
LMSYS Reputation Risk: Opaque Model Launch Concerns
By
–
Feels to me like a bit of a reputation risk to @lmsysorg though if this is indeed a stealth model launch They're supposed to be a neutral benchmarking tool, it's not a great look if they're working behind-the-scenes with model vendors in an opaque manner like this
-
Paper Shows AI Systems Lack Robust Adversarial Attack Protection
By
–
Their paper concludes with a note that this isn't a robust protection against adversarial attacks – more notes here
-
Universal AI Access: Either for Everyone or Nobody
By
–
It either works for everyone, or it works for no one.
-
LLM Knowledge Creation and Human Verification Requirements
By
–
For me part of the problem is that if an LLM did create "new knowledge" it would be incapable of verifying that what it had created was genuinely new – that's not possible without human involvement
-
AI Requires Domain Expertise: The Clinician Example
By
–
Working with Generative AI for knowledge work, over the past few years, one thing is very evident: you have to pair powerful A.I. with knowledgeable individuals, to make the most out of it. Trying to make a non-clinically trained person a clinician by aiding them with powerful