Skip to content

Frequently Asked Questions (FAQ)

This page contains answers to questions submitted via the AMA form. Updated regularly during and after the workshop.

General Questions

What if I'm not comfortable sharing certain data even in the sandbox?

Answer: That's perfectly valid! The sandbox is for practice - you can always use:

  • Hypothetical examples
  • Sanitized/anonymized data
  • Generic placeholders (e.g., "Polymer A", "[TEMP]°C")

The Red List protocol applies everywhere, including the sandbox, even though it's secure.

Can I use the frameworks (AUTOMAT/CO-STAR) with tools other than ChatGPT?

Answer: Absolutely! These frameworks work with any LLM:

  • Claude, Gemini, Copilot
  • Local models (Llama, Mistral)
  • Even specialized models (code assistants, research tools)

The principles are universal, clear structure, explicit constraints, defined outputs.

Technical Questions

Why does the sandbox model respond slower than ChatGPT?

Answer: Two main reasons:

  1. CPU vs GPU: ChatGPT runs on powerful GPUs; our sandbox uses CPU (slower but private)
  2. Model size: We use 3B parameters (fast, efficient) vs. ChatGPT's larger models

Trade-off: Privacy vs. speed. For training purposes, 3-8 second responses are fine!

Framework Questions

What if my prompt doesn't work even with frameworks?

Troubleshooting checklist:

  1. Is output format specified exactly? (table columns, bullet structure, etc.)
  2. Are constraints explicit? ("If missing, mark 'NR'" not just "summarize")
  3. Is the task clear? (avoid "help me with X" - be specific)
  4. Is context relevant? (remove unnecessary background)
  5. Is model capable? (use LLM for reasoning, NLM for extraction)

Post your prompt in AMA for personalized feedback!

Red List & Security

What if I accidentally shared something from the Red List?

Immediate steps:

  1. Stop using that conversation/session
  2. Notify your supervisor
  3. Document what was shared
  4. Follow company incident response

Prevention: When in doubt, use sandbox or don't share at all.

Application Questions

Can I use these techniques for non-materials-science work?

Answer: Yes! The frameworks are domain-agnostic:

  • AUTOMAT works for any structured task
  • CO-STAR works for any strategic communication
  • Red List applies to any company data
  • Green AI applies to any usage

Just adapt examples to your domain.

Workshop Logistics

Can I share these materials with colleagues?

Answer: Materials are proprietary to AmaDema employees.

You can:

  • Share within AmaDema (internal knowledge transfer)
  • Reference concepts in your work
  • Teach frameworks to AmaDema colleagues

You cannot:

  • Share materials externally (competitors, partners, public)
  • Post on public platforms
  • Distribute outside company

More Questions?

During workshop: Use the AMA form or ask instructor directly.

After workshop: Submit via AMA form - new answers added weekly.

Last updated: January 2026
Questions answered: Updated after each workshop session