All courses

Advanced · module A15

Frontier AI: AGI and Open Research Problems

By the end: Assess claims about AGI critically, explain the main open problems, and identify promising research directions

Start here
Covers
Where AI stands today, the unsolved problems, and the debates about general and superhuman intelligence
Contents
3 chapters, 16 topics
Time
About 3 hours
Level
Advanced
Your progress
0 / 16

This module builds on Large Language Models: Training, Adaptation and Evaluation. We suggest looking at it first, but the choice is yours.

Chapter 1 General Intelligence

  1. 1.1

    Defining Artificial General Intelligence

    Competing definitions and why they disagree

    Discussion seminar

    12 min Read
  2. 1.2

    Levels of AGI

    Performance, generality and autonomy as separate scales

    Exercise: place current systems on the levels

    10 min Read
  3. 1.3

    Measuring General Intelligence

    Turing-style tests, ARC-AGI and real-world task benchmarks

    Lab: attempt ARC-style puzzles

    12 minBeing written
  4. 1.4

    The Scaling Hypothesis and Its Critics

    Evidence for and against scaling as the path to AGI

    Debate

    12 minBeing written

Chapter 2 Open Research Problems

  1. 2.1

    Reasoning and Planning

    Where current models succeed and fail at multi-step reasoning

    Reading seminar

    12 minBeing written
  2. 2.2

    World Models and Embodied AI

    Learning how the world works, and AI in physical bodies

    Reading seminar

    12 minBeing written
  3. 2.3

    Robotics Foundation Models

    General-purpose robot learning from large data

    Case study

    10 minBeing written
  4. 2.4

    Continual and Lifelong Learning

    Learning new things without forgetting old ones

    Reading seminar

    10 minBeing written
  5. 2.5

    Causality in AI

    Moving from correlation to cause and effect

    Code: a simple causal analysis

    12 minBeing written
  6. 2.6

    Neuro-Symbolic AI

    Combining learning with logic and structured knowledge

    Reading seminar

    10 minBeing written
  7. 2.7

    AI for Science

    Protein structure, materials, mathematics and climate

    Case study: AlphaFold

    12 minBeing written

Chapter 3 The Long-Term Future

  1. 3.1

    Superintelligence and the Intelligence Explosion

    The arguments, and the strongest objections to them

    Debate

    12 min Read
  2. 3.2

    Machine Consciousness

    Philosophical positions, and what can and cannot be tested

    Discussion seminar

    10 minBeing written
  3. 3.3

    Future Computing for AI

    Quantum and neuromorphic computing: what they may and may not change

    Reading seminar

    10 minBeing written
  4. 3.4

    AI Timelines and Forecasting

    How to read predictions and their uncertainty

    Exercise: evaluate three forecasts

    10 minBeing written
  5. 3.5

    Module Review and Assessment

    Essay and a scored quiz

    Graded essay

    20 minBeing written