ProbAI Group Wiki
Welcome to the ProbAI group at Western CS!
Vision
Making AI agents capable of human-like decision-making.
Missions
- Equipping ML models with calibrated and cost-effective uncertainty-quantification capabilities.
- Creating efficient autonomous decision-making AI agents.
- Making the world a better place by accelerating scientific discovery through efficient AI systems.
Values
important
Our group values craft over volume.
- Few highly-cited, influental core works rather than 20 sloppy papers.
- Easy-to-use software libraries with proper, latest software engineering practices.
- Etc!
Code of Conduct
- Follow the group rules and culture.
- We are here to learn and work as a group. Getting along with everyone is the key---please be inclusive and collegial.
- Be professional.
- Be reliable, accountable, and on-time.
- Please skim all articles in this wiki at least once.
- You can then read them in more detail when you need them.
- Always check the wiki before asking.
- Contribute!
- We build our group together, so please contribute to the wiki.
- If you learned something new about the system or have new tips, please write them down.
- Simply clone the repo (https://github.com/kristiadigroup/group-wiki) and submit a pull request. Once Agustinus approves the change, you can merge yourself.
LLM Code of Conduct
We value learning and self- and group-improvement. Thus, LLMs are only acceptable for:
- "Study Mode". Turn this feature on explicitly.
- Even then, whiteboard discussion with fellow students and with Agustinus are much preferred.
- Coding Buddy.
- E.g., asking "how to partition an array in NumPy".
- Not the same as agentic/vibe coding. Not the same as Copilot/Cursor tab autocomplete (turn this off).
- You should definitely learn AI-enhanced and agentic coding (industry demands), but in your free time, for side projects.
- Grammar correction.
- Even then, good-old Grammarly (free tier) is preferred. Just fix the red errors in Grammarly. Don't use the rephrase function.
Taking shortcut now = work many times harder in the future. Also, high potential of embarrassment. Our group treats other LLM usages as misconduct.