INF1-IP: Week 5
Lecture 8 - Introduction to Generative AI
This lecture does not introduce new programming concepts and as such is not assessed and has no related self-guided notebook or key terms. We give this lecture because we think it is important for programmers to have a basic understanding of the field of AI, and give it in this week to not provide too much new programming material in the week of the first Class Test.
- Lecture slides: TBA
Learning outcomes
- Describing the field of artificial intelligence and its major subfields
- Understanding, on a high level, how large language models generate text
- Understanding the ethical implications of creating and using LLMs and the causes of these issues
Further reading
- The University of Edinburgh's generative AI platform: ELM
- Anatomy of an AI System, 2018, by Kate Crawford and Vladan Joler: shows everything that goes into building an Amazon Echo device! Pre-GenAI but still very relevant today.
- Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT '21). Association for Computing Machinery, New York, NY, USA, 610–623. https://doi.org/10.1145/3442188.3445922
Lecture 9 - Code quality 1: Testing and documentation
- Lecture slides: TBA
- Self-guided notebook (right-click → Save As... then remove the .txt part so it is saved as an .ipynb)
Learning outcomes
- Applying decomposition and DRY to give functions clear, reusable jobs
- Explaining test-driven development and the red → green → refactor cycle
- Writing
assertchecks to automatically verify program functionality
Key terms
Click to expand.
Decomposition
Breaking a problem into smaller parts with clear responsibilities.
DRY
“Don't Repeat Yourself”: keeping a shared rule in one place.
Refactoring
Improving code's structure while preserving its intended behaviour.
Assertion
A condition that is checked during execution. A false condition causes an
AssertionError.Test case
An input together with its expected result or behaviour.
Test-driven development
An approach to coding in which you write a failing test for a small requirement, implement it, and then improve structure while keeping the tests passing.
Regression test
A retained check that can catch the return of a repaired bug.
New syntax
Click to expand.
assert conditionContinues silently if
conditionevaluates toTrueor stops the program withAssertionErrorif not.assert condition, "message"Includes a message with the
AssertionErrorif the assertion fails.AssertionErrorA type of error raised when an assertion fails.
Further reading and practice
- Official Python documentation: The
assertstatement. - Python's style guide: see the naming conventions for classes, functions, and variables.
- W3Schools:
assertkeyword. - Helsinki Python MOOC: Common error messages.
- Martin Fowler: Test Driven Development - a quick article that introduces the philosophy.
Tutorial
Tutorial 3: The Tools of the Trade 2 walks you through installing packages from the internet and setting up virtual environments to manage more advanced Python projects.
- Interactive notebook (right-click → Save As... then remove the .txt part so it is saved as an .ipynb)
- Note that this notebook was created to be used in Notable; its interactive checker cells might not work in every environment.
Assessment
Class Test 1 will take place on Thursday 22nd October in AT6:06. Check your timetable to see which session you are scheduled for.
For details, see Assessment.