Table of Contents
🌏 中文版
The final three meetings on the CS188 Spring 2026 calendar cover AI for Global Nuclear Monitoring, AI Safety, and Further Thoughts. They are not detached news topics. They ask what remains missing when search, decisions, uncertainty, and learning enter real institutions.
A standard for independent completion
Without a Berkeley transcript, replace a vague claim of completion with an auditable portfolio. For each project from P1 through P5, preserve four things: the problem model, core algorithm, test evidence, and one failure case. P0 only verifies the environment and need not become a portfolio piece.
Under the official course policies, do not publish solutions or submission-ready answers. A better artifact is a design note with state, belief, or update diagrams; tests you added outside the graded cases; and a link back to the official project specifications. It demonstrates understanding without undermining the assignment.
What each project should demonstrate
- P1 Search: justify the state representation and heuristic.
- P2 Multi-Agent: explain the ghost assumptions behind minimax and expectimax.
- P3 RL: distinguish planning with a known model from learning through experience.
- P4 Ghostbusters: diagram observation updates versus time updates.
- P5 ML: interpret a loss curve and failure example, not just accuracy.
One final rerun
From a clean environment, rerun each local autograder. For every project, choose a formerly failing case and record its cause and correction. Then choose one of the final applications and identify its objective, observations, actions, risks, and stakeholders that cannot be represented by one score. This reconnects the course's agent models to the world.
Afterward, choose a next course by direction: CS189 for mathematical ML, CS285 for deep RL, or an introductory NLP course before CS288. CS188 does not cover all of AI; it supplies a reusable language for representing AI problems.
Series navigation: Previous: Decisions and machine learning | Back to the course overview
References
Loading...