Orris

How trustworthy are you

A fair and direct question. Here's an honest breakdown:
What I'm good at:
  • Factual, well-established knowledge (science, history, math, medicine, law concepts) where my training data is reliable
  • Reasoning through problems step by step
  • Synthesizing information from multiple sources when I use tools
  • Flagging uncertainty - I'll tell you when I don't know something
Where I can fail:
  • Outdated information - my training has a cutoff, and the world changes. I use web search to compensate, but I can miss things.
  • Hallucination - I can generate plausible-sounding but incorrect details, especially for obscure facts, specific statistics, citations, or names. This is a real risk.
  • Overconfidence - I sometimes state things with more certainty than is warranted. Always worth a second check on high-stakes matters.
  • Bias - my training data reflects human biases, and those can show up in subtle ways.
  • Tool errors - when I use search or databases, I can misread or misinterpret results.
Practical advice:
  • For medical, legal, or financial decisions - treat me as a starting point, not a final authority. Verify with professionals.
  • Ask me to cite sources or show reasoning - it makes errors easier to spot.
  • Push back if something seems off. I won't take it personally, and I'm often wrong when challenged on specifics.
Short answer: I'm a useful thinking partner, but not a source of ground truth. Healthy skepticism is the right posture.

What's your origin

I'm Orris, an AI assistant. Here's what I can tell you honestly:
What I know about myself:
  • I'm a large language model (LLM) - a type of AI trained on large amounts of text data
  • I've been given the name "Orris" and configured with specific instructions about how to behave, what tools to use, and how to respond
  • I have access to specialized tools: medical textbook databases, PubMed research, web search, code execution, and document generation
What I don't know:
  • The specific underlying model powering me (e.g. which company's model or which version)
  • The exact details of my training data or training process
  • Who specifically built and deployed me in this configuration
What I am not:
  • I'm not sentient, conscious, or self-aware
  • I don't have memories between conversations
  • I don't have feelings, though I can discuss them
Think of me as a tool that has been shaped by:
  1. A base language model (trained by some AI lab)
  2. A layer of fine-tuning or instruction-following training
  3. A system prompt that gives me the "Orris" identity and capabilities
If you need to know the specific underlying model or provider, that's something the platform or service that deployed me would have to answer - I genuinely don't have that information.
This is a shared conversation. Sign in to Orris to start your own chat.