**Study Group Topic: Constitutional AI**
Dive into the foundational research on using principles rather than human feedback to train safe AI.

**Constitutional AI: Harmlessness from AI Feedback** (Bai et al., 2022)
The original paper introducing Constitutional AI - training AI assistants to be helpful and harmless using self-critique and AI-generated feedback rather than human labels.
🔗 [https://arxiv.org/abs/2212.08073](https://arxiv.org/abs/2212.08073)

**Collective Constitutional AI: Aligning a Language** Model with Public Input (Huang et al., 2024)
Extends Constitutional AI by incorporating \~1,000 Americans' input to democratically create principles for AI behavior.
🔗 [https://arxiv.org/abs/2406.07814](https://arxiv.org/abs/2406.07814)

**📚 Full reading list and How-To:** [https://github.com/suzana-ilic/study_model_behavior](https://github.com/suzana-ilic/study_model_behavior)

Join us for our monthly reading group where we dive into the research and specs that shape how AI systems like ChatGPT and Claude actually behave. We read together for 30 minutes, then discuss for 30 minutes. Pre-reading is recommended, but not required.

**Who is this for?**
Anyone curious about how AI systems work—researchers, builders, policy folks, or just thoughtful people who use these tools and want to understand them better. No technical background needed. We start with accessible industry standards and papers and build from there.

**What will we read?**
We're working through resources in six areas:

1. Industry Specs — How leading AI companies define model behavior
2. Constitutional AI — Training models with principles instead of human feedback
3. Safety Methods — RLHF and alignment techniques
4. Behavioral Science — How researchers study what AI actually does
5. Interpretability — Understanding what's happening inside the models
6. Critical Perspectives — Challenges to current approaches

**Format**

* 30 minutes: Read together (with discussion questions)
* 30 minutes: Talk through key insights and implications
* Monthly sessions

**Location:** [Online]
💻 RSVP for Zoom Link
⚙️ Discord [https://discord.gg/CT7nBdYCsY](https://discord.gg/CT7nBdYCsY)
📬 Updates: [https://mltaicommunities.substack.com/](https://mltaicommunities.substack.com/)

Machine Learning Tokyo

Suzana

Ramdhan Hidayat

pavansai tadepalli

Dimitrios Dimopoulos

gretel

Jerico Delos Reyes

Pradeep

Himanshu

Adriane Mikko Amorado

Akie Takeuchi

MARIANO DOMINGUEZ

Yasuhiro Okazaki

Jun Fujisawa

Chan Chong Wei

Amee Kristin Ayco

Kamalavani Thota

Jeffrey Seifert

Stephany

Chaitanya Sai Nutakki

Model Behavior Study Group: Constitutional AI

Online event

Artificial_Intelligence

후원사

그룹

{PRO_NETWORK_NAME}은(는) Meetup 프로로서 전 세계 회원 수가 {PRO_MEMBER_COUNT}명 이상이며 Meetup이 {PRO_GROUP_COUNT}개에 달합니다. Meetup 프로는 사용자, 파트너, 기여자, 회원의 네트워크를 조직하고 소통할 수 있게 해주는 전문적인 툴입니다.

MLT & AI Communities. Join us on https://luma.com/AI-Communities

You're hiring? Drop your job description, get your shortlist in minutes.

Corehire AI

Artificial_Intelligence

소개

후원사

Corehire AI

예정된 이벤트

소개

후원사

Corehire AI

그룹