Artificial Intelligence
by Pieter Abbeel · edX
Our Verdict
Worth it — with caveatsUC Berkeley's CS188 'Artificial Intelligence' (the edX MOOC by Dan Klein and Pieter Abbeel) is widely regarded as one of the best AI courses ever put online, and for the right learner it is genuinely worth it for its rigorous, algorithm-first treatment of classical AI built around the famous Pac-Man programming projects. The catch is logistical: the original edX MOOC is now archived rather than offered as a live, regularly enrolling course, and the same complete material (lectures, slides, homeworks, and all Pac-Man projects) is freely and legally available at ai.berkeley.edu. Scope matters too: the MOOC/first-half covers search, constraint satisfaction, adversarial game-playing, Markov decision processes, and reinforcement learning, but it is NOT a modern deep-learning or generative-AI course and largely omits neural networks. It demands a real CS background (data structures, comfortable Python, and probability) and is frequently described as one of the harder offerings of its kind. Treat the catalog's '$199' verified-certificate price with caution: I could not confirm a currently purchasable certificate at the listed edX URL, which returned 404 at review time.
The content is elite and the Pac-Man projects are outstanding, but the original edX CS188.1x MOOC is archived (no active live enrollment) and the identical curriculum is free at ai.berkeley.edu, so paying $199 for a verified certificate is hard to justify and could not be verified as currently available. It is a strong 'take' for learners who want classical/symbolic AI foundations and have the prerequisites; it is the wrong pick for anyone expecting modern deep learning, LLMs, or a beginner-friendly on-ramp.
Best for: CS students and self-taught engineers who already know data structures, can write Python comfortably, and have had a first course in probability, and who want a rigorous grounding in classical AI: state-space search and heuristics, constraint satisfaction, minimax/expectimax game-playing, MDPs, and reinforcement learning. It is ideal for people who learn best by implementing algorithms from scratch (the Pac-Man projects) rather than watching demos, and for those preparing for AI fundamentals interviews or further study in robotics and decision-making.
Skip if: Beginners with no programming or math background, and anyone whose goal is modern deep learning, neural networks, computer vision, NLP, or generative AI / LLMs - this course predates and largely excludes that material. Learners who need structured deadlines, graded support, an instructor/community, or a recognized paid certificate should also look elsewhere, since the offering today is effectively self-paced archived material.
About This Course
Berkeley's upper-division AI course covering search, constraint satisfaction, game playing, MDPs, and reinforcement learning.
What You'll Learn
Curriculum
Uninformed and informed state-space search: depth-first, breadth-first, uniform-cost, and A* search with admissible/consistent heuristics. Reinforced by Project 1 (Search) in the Pac-Man framework.
Representing problems as CSPs and solving them with backtracking, filtering, and related techniques.
Minimax and expectimax for multi-agent games, alpha-beta pruning, utilities, and evaluation-function design. Reinforced by the Multi-Agent Pac-Man project.
Modeling sequential decision-making under uncertainty; computing optimal policies with value iteration and policy iteration.
Learning to act from experience: model-based and model-free methods, Q-learning, and approximate Q-learning. Reinforced by the Reinforcement Learning project on gridworld, Pac-Man, and a simulated crawler.
Prerequisites
- A first course in algorithms and solid familiarity with basic data structures (Berkeley on-campus prereqs are CS61A/CS61B)
- Comfortable Python programming - projects are in Python and you are expected to ramp up quickly
- A first course in probability and comfort with mathematical notation (Berkeley on-campus prereq is CS70)
- Some linear algebra and calculus is helpful, though the small amount needed is introduced as required
Instructor
Pieter Abbeel
Instructor · edX
Pros & Cons
Pros
- The Pac-Man programming projects are frequently cited as among the best assignments in any online course - you implement real AI algorithms and watch them run, which makes abstract material concrete and memorable
- Authoritative, rigorous instruction from leading Berkeley AI faculty (Dan Klein and Pieter Abbeel), covering the same material as the on-campus CS188
- All course materials - lecture videos, slides, homeworks, exams, and the full Pac-Man project specs - are freely and legally available at ai.berkeley.edu, so the core learning costs nothing
- Strong, durable foundation in classical/decision-theoretic AI (search, CSPs, games, MDPs, RL) that underpins robotics and sequential decision-making and remains relevant despite the deep-learning era
Cons
- Steep difficulty and a high prerequisite bar - reviewers repeatedly note you need solid Python, algorithms-from-pseudocode skill (including recursion), and probability, making it unsuitable for true beginners
- The edX MOOC is archived rather than a live, regularly enrolling course, and historically covered only the first half of CS188 - Bayes nets, HMMs/particle filtering, and the machine-learning/classification material were a separate part that was never fully maintained as a MOOC
- Not a modern course: it largely omits deep learning, neural networks, computer vision, NLP, and generative AI / LLMs
- Largely self-directed with no real instructor or cohort support, which several learners described as isolating; the catalog '$199' verified certificate could not be confirmed as currently purchasable (the listed edX URL returned 404)
Alternatives To Consider
Frequently Asked Questions
Is Artificial Intelligence free?
Artificial Intelligence is $199. The core course is effectively free: full lectures, slides, homeworks, and all Pac-Man projects are openly available at ai.berkeley.edu, and the edX version offered free audit. The catalog lists a $199 verified certificate, but the linked edX course page returned 404 at review time and the MOOC appears archived, so do NOT assume a paid certificate is currently available - verify on edX before expecting to pay for one.
Who is Artificial Intelligence for?
CS students and self-taught engineers who already know data structures, can write Python comfortably, and have had a first course in probability, and who want a rigorous grounding in classical AI: state-space search and heuristics, constraint satisfaction, minimax/expectimax game-playing, MDPs, and reinforcement learning. It is ideal for people who learn best by implementing algorithms from scratch (the Pac-Man projects) rather than watching demos, and for those preparing for AI fundamentals interviews or further study in robotics and decision-making.
What will you learn in Artificial Intelligence?
Formulate problems as state-space search and implement DFS, BFS, uniform-cost search, and A* with admissible heuristics; Model and solve constraint satisfaction problems (CSPs); Build adversarial game-playing agents using minimax and expectimax, with alpha-beta pruning and hand-designed evaluation functions; Define and solve Markov Decision Processes (MDPs) via value iteration and policy iteration.
What are the prerequisites for Artificial Intelligence?
A first course in algorithms and solid familiarity with basic data structures (Berkeley on-campus prereqs are CS61A/CS61B); Comfortable Python programming - projects are in Python and you are expected to ramp up quickly; A first course in probability and comfort with mathematical notation (Berkeley on-campus prereq is CS70); Some linear algebra and calculus is helpful, though the small amount needed is introduced as required.
Is Artificial Intelligence worth it?
The content is elite and the Pac-Man projects are outstanding, but the original edX CS188.1x MOOC is archived (no active live enrollment) and the identical curriculum is free at ai.berkeley.edu, so paying $199 for a verified certificate is hard to justify and could not be verified as currently available. It is a strong 'take' for learners who want classical/symbolic AI foundations and have the prerequisites; it is the wrong pick for anyone expecting modern deep learning, LLMs, or a beginner-friendly on-ramp.
How we reviewed this course
This is an independent editorial assessment by Cursarium, based on edX's published course materials and aggregated public learner feedback (last reviewed 2026-06). We have not independently completed the course. Links to providers are standard references, not paid placements.
Sources
- Official edX catalog URL for the course (returned HTTP 404 at review time, indicating the listing is stale/archived)
- ai.berkeley.edu - official UC Berkeley CS188 materials hub (free lectures, slides, homeworks, Pac-Man projects)
- UC Berkeley CS188 Fall 2024 course site (authoritative current syllabus: search, CSPs, games, MDPs, RL, Bayes nets, HMMs, ML)
- The Pac-Man Projects (modelai.gettysburg.edu) - official project names and the exact algorithms each one implements
- Class Central - CS188.1x course page and editorial review ('one of the best MOOCs on the web'; first-half scope; archived status)
- Class Central report: Course Review: Artificial Intelligence by UC Berkeley on edX (strengths, Pac-Man projects, isolation, prerequisites)
- GetYourEducation listing - third-party rating (4.8 from 30 votes), duration, cost, certificate, instructors
- UC Berkeley CS188 Fall 2014 course information - explicit prerequisites (CS61A/61B, CS70/probability, Python)