Instructor: Dan Ryan
Human and Machine Intelligence Alignment
Fall 2025
Welcome and Getting Started
Agenda
- Who am I?
- Why are we here? What is alignment?
- Who are you?
- What will we do?
Ryan: Human and Machine Intelligence Alignment Fall 2025
Instructor
Ryan: Human and Machine Intelligence Alignment Fall 2025
Dan Ryan
Ask Me Anything*
Ryan: Human and Machine Intelligence Alignment Fall 2025
*but not about the course, that's what we'll talk about next.
Why Are We Here?
Ryan: Human and Machine Intelligence Alignment Fall 2025
2015
Do YOU think we should worry about intelligent machines that do exactly what we ask them to do?
Ryan: Human and Machine Intelligence Alignment Fall 2025
Ryan: Human and Machine Intelligence Alignment Fall 2025
Who are you?
Ryan: Human and Machine Intelligence Alignment Fall 2025
Ryan: Human and Machine Intelligence Alignment Fall 2025
2019
What
Is
Value
Alignment?
Ryan: Human and Machine Intelligence Alignment Fall 2025
2019
Ryan: Human and Machine Intelligence Alignment Fall 2025
Dan Ryan on Social Theory and Alignment [6m]
This
is not our
first rodeo
What This Course Is NOT
Ryan: Human and Machine Intelligence Alignment Fall 2025
Introduction to AI
Computer Science Ethics
Ethics of AI
AI Alignment (CS Version)
What This Course IS: Four Takes
Ryan: Human and Machine Intelligence Alignment Fall 2025
Ryan: Human and Machine Intelligence Alignment Fall 2025
The Four Intelligence Alignments
Ryan: Human and Machine Intelligence Alignment Fall 2025
Living with Other Human Intelligences
Living with Organizational Intelligences
Living with Machine Intelligences
Living with Expert Intelligences
Ryan: Human and Machine Intelligence Alignment Fall 2025
AI 101
Alignment 101
How to read hard stuff
Thinking Analogically
The Four
Alignment
Problems
Isn't It Just "Ethics"?
Principles Are
Not Enough!
Align is a Verb
Shared Meaning
Hierarchy
Groups
Markets
Qualification
Records
Control
Deterrence
Incentives
Safety
Institutions
Foundations
Humans
Have Been Doing Alignment Forever
Traits Are
Not Enough!
Humans +
Humans
Principals + Agents
Normies +
Magicians
Humans +
Machines
Humans
Organizations
Experts
Machines
Humans
Organizations
Experts
Machines
Humans
Organizations
Experts
Machines
Humans
Organizations
Experts
Machines
Humans
Organizations
Experts
Machines
Humans
Organizations
Experts
Machines
Humans
Organizations
Experts
Machines
Basic Requirements
Ryan: Human and Machine Intelligence Alignment Fall 2025
27 class meetings.
~25 pre-class work.
<25 post-class work.
Multiple draft essay.
Alignment card deck.
Oral exam.
Alignment Cards
Ryan: Human and Machine Intelligence Alignment Fall 2025
Alignment Chat
Ryan: Human and Machine Intelligence Alignment Fall 2025
Next Time on...
PostClass Work
- Play with our Alignment Guide GPT, Lumen
PreClass Work AI Bootcamp
- About 25 minutes of video and skim ebook Machine Learning for Humans
Ryan: Human and Machine Intelligence Alignment Fall 2025
Further Reading on Alignment
-
IBM: "What is alignment?"
- General answer with examples of things IBM is developing
-
Wikipedia. "Al Alignment" 25-30 min read
- Read first three sections and skim the rest for a first reading
-
Vinod Chugani 2024: "What is AI Alignment? Ensuring AI Works for Humanity" on DataCamp blog · 12 min read
- Ignore opening idea (building in values); after that several good break downs of concepts related to alignment
-
Paul Christiano. 2018. "Clarifying 'AI alignment'." 4 min read
- head of AI safety at the US AI Safety Institute
-
Ji, J. et al. 2024. "AI Alignment: A Comprehensive Survey" 60PP!
- dense text, read from the outside in - look at contents and overall shape
-
Tech AI Digest. 2024. "AI Alignment Explained: Ensuring AI Follows Human Values" (11min podcast)
- 2 person conversation - very beginner friendly
Ryan: Human and Machine Intelligence Alignment Fall 2025
We just learned three terms
maximize a utility function
inverse reinforcement learning
value alignment
Ryan: Human and Machine Intelligence Alignment Fall 2025