1/99
Looks like no tags are added yet.
[12 AI foundations] What does computer vision work with?
Images and video.
[12 AI foundations] What does speech recognition convert?
Spoken audio into text.
[12 AI foundations] What does natural language processing work with?
Human language.
[12 AI foundations] What does generative AI produce?
New content, such as text, images, or code.
[12 AI foundations] What year was the Dartmouth AI workshop?
1956.
[12 AI foundations] What year was ELIZA introduced?
1966.
[12 AI foundations] What was ELIZA?
An early conversation program using scripted patterns.
[12 AI foundations] What is the ELIZA effect?
Attributing understanding or empathy to a program that lacks it.
[12 AI foundations] What is an AI winter?
Reduced enthusiasm and funding after unmet expectations.
[12 AI foundations] How does the study guide characterize AI history?
Repeated booms and winters.
[12 AI foundations] Why did hand-written expert-system rules struggle?
Exceptions made them brittle and difficult to maintain.
[12 AI foundations] How does machine learning differ from hand-written rules?
It learns patterns from examples.
[12 AI foundations] What ImageNet breakthrough year does the guide emphasize?
2012.
[12 AI foundations] What architecture introduced in 2017 helped language models scale?
The transformer.
[12 AI foundations] What year was ChatGPT released?
2022.
[12 AI foundations] What are features in machine learning?
Input variables.
[12 AI foundations] What are labels in supervised learning?
Known target outcomes.
[12 AI foundations] What is training?
Adjusting a model using examples and errors.
[12 AI foundations] What is a prediction?
A model's output for an input.
[12 AI foundations] What is the basic learning loop?
Examples, prediction, comparison, adjustment, repeat.
[12 AI foundations] What shortcut might a husky-versus-wolf classifier learn?
Snow in the background instead of animal features.
[12 AI foundations] What does that snow example illustrate?
Models can learn unintended correlations.
[12 AI foundations] How are images, speech, and text represented for machine learning?
As numbers.
[12 AI foundations] What is an LLM's basic next-token objective?
Predict the next token from preceding context.
[12 AI foundations] Does fluent AI output guarantee factual accuracy?
No.
[13 UX and UI] What does UX stand for?
User experience.
[13 UX and UI] What does UI stand for?
User interface.
[13 UX and UI] What does UX design focus on?
The overall experience of accomplishing a user's goal.
[13 UX and UI] What does UI design focus on?
The interface's appearance and controls.
[13 UX and UI] According to the lecture, do users usually read or scan pages?
Scan.
[13 UX and UI] What does satisficing mean?
Choosing an option that seems good enough.
[13 UX and UI] What does muddling through mean in website use?
Trying things without fully understanding the system.
[13 UX and UI] What is Krug's first law of usability?
Don't make me think.
[13 UX and UI] What should a usable page make obvious?
Where to begin and how to act.
[13 UX and UI] What does visual hierarchy communicate?
Which information is most important.
[13 UX and UI] Which scanning pattern suits text-heavy pages?
The F pattern.
[13 UX and UI] Which scanning pattern suits less text-heavy layouts?
The Z pattern.
[13 UX and UI] Why use familiar interface conventions?
Users already know what they mean.
[13 UX and UI] Why make clickable elements obvious?
So users can recognize available actions.
[13 UX and UI] What is visual noise?
Clutter that competes with useful information.
[13 UX and UI] What is a user persona?
A research-based fictional representative of a user type.
[13 UX and UI] What does a user story describe?
A user's goal and why it matters.
[13 UX and UI] What does a sitemap organize?
Pages and navigation.
[13 UX and UI] What does a wireframe show?
Basic structure and content placement.
[13 UX and UI] What does a prototype let you test?
Proposed interactions and user flows.
[13 UX and UI] Why analyze UX metrics?
To check whether users can complete intended goals.
[13 UX and UI] What should guide UX improvements?
User research, testing, and observed behavior.
[14 Prompting and web apps] What is a prompt?
Input that guides a generative model's output.
[14 Prompting and web apps] What is the prompt-engineering loop?
Prompt, generate, evaluate, modify.
[14 Prompting and web apps] What is a directive in a prompt?
The main instruction or question.
[14 Prompting and web apps] What do prompt examples demonstrate?
The desired pattern or output.
[14 Prompting and web apps] What do output-format instructions specify?
The shape or structure of the answer.
[14 Prompting and web apps] What do style instructions specify?
Tone, look, or genre.
[14 Prompting and web apps] What does a role instruction assign?
A persona or perspective.
[14 Prompting and web apps] Why include additional information in a prompt?
To provide facts the model needs.
[14 Prompting and web apps] What is zero-shot prompting?
Instructions without examples.
[14 Prompting and web apps] What is few-shot prompting?
Instructions with examples.
[14 Prompting and web apps] What is thought-generation prompting?
Encouraging reasoning or planning before an answer.
[14 Prompting and web apps] What is decomposition?
Breaking a task into smaller subtasks.
[14 Prompting and web apps] What is ensembling?
Combining multiple generated answers.
[14 Prompting and web apps] What is self-criticism prompting?
Having the model review and revise its output.
[14 Prompting and web apps] What is prompt sensitivity?
Output changes caused by small prompt changes.
[14 Prompting and web apps] What is sycophancy?
Drifting toward the user's stated opinion.
[14 Prompting and web apps] What is overconfidence?
Expressing more certainty than accuracy warrants.
[14 Prompting and web apps] What is prompt injection?
Untrusted content attempts to redirect the system's instructions.
[14 Prompting and web apps] What is jailbreaking?
Trying to induce behavior the model is designed to refuse.
[14 Prompting and web apps] What is package hallucination?
Generated code names a nonexistent software package.
[14 Prompting and web apps] Why can package hallucination be dangerous?
Attackers may publish malicious packages under those names.
[14 Prompting and web apps] What does an app prompt's context explain?
What the app is and who it serves.
[14 Prompting and web apps] What do technical constraints specify?
How the app must be built and run.
[14 Prompting and web apps] What should an app prompt's features describe?
Specific actions and behaviors.
[14 Prompting and web apps] Why include sample data in an app prompt?
To make the intended content concrete.
[14 Prompting and web apps] What does look and feel specify?
The app's visual style and layout.
[14 Prompting and web apps] What does HTML provide?
Content and structure.
[14 Prompting and web apps] What does CSS control?
Appearance and layout.
[14 Prompting and web apps] What does JavaScript implement?
Behavior and interactions.
[14 Prompting and web apps] What does no server mean for the class app?
It runs locally without a backend.
[14 Prompting and web apps] What does in memory mean for app data?
Stored in variables while the page is open.
[14 Prompting and web apps] What happens to in-memory app data on refresh?
It disappears.
[14 Prompting and web apps] What does localStorage do differently?
Keeps data across refreshes.
[14 Prompting and web apps] How should you check whether generated app code works?
Run it and test the behavior.
[14 Prompting and web apps] What makes a useful app-repair prompt?
One concrete defect and the desired behavior.
[15 AI economics] What is Jevons paradox?
Greater efficiency can increase total demand.
[15 AI economics] How might cheaper scan interpretation affect scan demand?
It may increase the number of scans requested.
[15 AI economics] Does automating a task necessarily eliminate the whole job?
No.
[15 AI economics] What is codified knowledge?
Explicit knowledge that can be written down.
[15 AI economics] What is tacit knowledge?
Experience-based knowledge that is harder to articulate.
[15 AI economics] Why may entry-level tasks face AI exposure?
Many rely on codified knowledge.
[15 AI economics] How can AI complement experienced workers?
It can support their judgment and contextual knowledge.
[15 AI economics] Does a faster individual task guarantee firm-wide profit gains?
No.
[15 AI economics] What can delay firm-wide AI productivity gains?
Integration, bottlenecks, review, training, and workflow changes.
[15 AI economics] What does the factory-electrification example emphasize?
New technology often needs organizational redesign.
[15 AI economics] What is approval-based AI deployment?
A human approves actions before execution.
[15 AI economics] What is escalation-based AI deployment?
AI handles routine cases and sends exceptions to humans.
[15 AI economics] What is agentic deployment?
AI carries out multi-step tasks with greater autonomy.
[15 AI economics] What is the circular-financing concern in AI?
Funding and spending commitments circulate among connected AI businesses.
[15 AI economics] What is model training's basic cost pattern?
Large upfront resource costs.
[15 AI economics] What is inference?
Running a trained model to produce outputs.
[15 AI economics] Does inference create continuing operating costs?
Yes.
[15 AI economics] Does usage growth alone prove AI-lab profitability?
No.