🌏 中文版
Video status: Videos included. Source details
Lecture 7 of MIT 6.S191 2026 is The Three Laws of AI: Safety Through Observability and Evaluation. It Starts from Asimov’s literary laws and examines modern safety protocols through traces, test data, and continuous evaluation. This note uses only the official 2026 slides and video; it does not mix in similarly named material from 2025.
What to take away
- See why abstract safety principles do not become guarantees by themselves
- Treat observability as a prerequisite for investigation and improvement
- Use a fixed test set to detect regressions after model or prompt changes
These goals have one thing in common: recognizing terminology is insufficient. You should be able to identify inputs, outputs, the learning signal, and the main constraint before moving on.
A safety principle must become a policy, observable events, test cases, and alert thresholds before it becomes an operational control. Traces show what a system did, and a fixed evaluation set reveals regressions; neither proves safety across every unknown situation.
How to watch
First scan the sections and diagrams in the official slides, then watch the official video. On a second pass, pause at equations and architecture diagrams and redraw them in your own notation. Afterward, close the material and write three central ideas plus one unresolved question.
An exercise for tonight
Save ten representative cases with inputs, outputs, and version metadata, then rerun them after the next update.
“Finished” means leaving a checkable diagram, calculation, program output, or short note—not merely reaching the end of the video. You should also be able to explain one failure mode to someone else.
Scope and limits
6.S191 is a high-intensity introduction, and this article is only a lecture guide. It does not replace the full recording, rigorous derivations, or instructor feedback. Use a semester course or primary papers when a topic needs theoretical depth.
Course video sources
Rechecked against the live official course page on 2026-10-10: the lecture numbers and recording links match and the videos are public and embeddable.
Original videos: MIT 6.S191: The Three Laws of AI
Course and recording entries:
Checked: 2026-10-10.
Content check: verified against the video transcript (2026-10-10): this is a guest lecture by Doug Blank, head of research at Comet, not one of the main instructors. The transcript starts from Asimov's three laws of robotics, then uses Comet's Opik to demo system-prompt jailbreaks, dataset-plus-metric evaluation experiments (including LLM-as-judge), trace logging and agent tool calls, and ends with his version of modern AI rules (log traces, grow a test dataset incrementally, evaluate continually, publish results), matching this note's trace, fixed test set and continuous evaluation.
Update Log
- 2026-10-10: Added explicit video status and checked recording sources and access notes.
- 2026-10-10: Rechecked video status. Live-checked the official course page: lecture numbers and recording links match and the videos are public, so the status is now “Videos included.”
- 2026-10-10: Checked the video content against its transcript. Confirmed this is a guest lecture by Doug Blank (Comet) and the content matches this note.
References
Loading...