AI Foundations for Everyone
How AI actually works, from the prediction engine underneath to the agents acting on your behalf. One course from first principles to real work, with the landmark studies taught as Spotlight lessons of their own.
13 sections ยท 42 lessons
Course outline
What Happens When It Answers
- There Is No Answer Database
- Picking From The Spread
- Decide Whether You Want Variety
How It Learned Everything It Knows
- Where The Patterns Came From
- Billions Of Dials
- The Scale Of The Bet
From Predictor To Assistant
- A Predictor Is Not An Assistant
- Taught By Preference
- Why It Stops At A Date
Memory And Context
- Working Memory, Not Storage
- The Middle Gets Lost
Inside The Machine
- Words Become Directions
- Attention Mixes In The Context
- Back To Probabilities
- Spotlight: The Paper That Started It
The Jagged Frontier
- The Boundary Is Not Where You Think
- Sort Your Own Tasks
- Spotlight: Centaurs And Cyborgs
Why It Makes Things Up
- Guessing Was The Winning Strategy
- Where To Expect It
- Spotlight: The Arithmetic Of Bluffing
The Quiet Failures
- Numbers Are Just Text To It
- Make It Show Its Working
- It Was Trained To Please You
- How To Get A Real Critique
Getting The Output You Want
- It Cannot Infer What You Did Not Write
- Show, Don't Tell
- Two Very Different Modes
- Grounded Is Not Guaranteed
- The Shape Is Part Of The Request
- Not A Ladder
- Buying Time To Think
Checking The Work
- Reading Well Is Not Being Right
- Match The Checking To The Stakes
- Spotlight: Feeling Faster, Measured Slower
Handing It Over
- Parts, Not Whole Tasks
- What Never Gets Handed Over
- Spotlight: The Cost Of Looking Finished
Agents
- The Loop
- Errors Compound
- Where The Brakes Go
Staying Current
- What Churns And What Holds