E077 Ch1 – WGU Lessons 1.1–1.3 + Percipio
Call Kai
Learn
Practice Test
Spaced Repetition
Match
Flashcards
Knowt Play1/54
There's no tags or description
Looks like no tags are added yet.
Name | Mastery | Learn | Test | Matching | Spaced | Call with Kai | Chat |
|---|
No analytics yet
Send a link to your students to track their progress
55 Terms
Artificial intelligence (AI) — WGU definition
Technologies that enable computers to perform tasks that typically require human intelligence, such as recognizing patterns, making predictions, and analyzing information.
AI engineering
A discipline focused on designing, deploying, and maintaining systems that integrate AI models into real-world applications.
AI system
A software system that uses data and algorithms to generate predictions, recommendations, or decisions as part of an application or service.
Data science
A field that focuses on analyzing data, identifying patterns, and building models that generate insights or predictions from datasets.
AI research
The study and development of new algorithms, methods, and theoretical approaches that advance AI capabilities.
Production system
An operational software environment where applications and services are deployed for real users and must function reliably at scale.
System ownership
Responsibility for maintaining and managing a system after deployment, including monitoring, updates, and operational response.
Data pipeline
A series of processes that collect, transform, and deliver data so it can be used by applications or models.
Scalability
The ability of a system to handle increased data volume or user demand without losing performance.
Reliability
A system's ability to function consistently and produce correct results over time.
Scenario: An online store wants AI product recommendations. What does the AI engineer do (not the researcher or data scientist)?
Integrates the trained model into the storefront, connects it to data pipelines and the user interface, makes sure it runs reliably at scale, and owns ongoing operation by monitoring and fixing issues.
Scenario: A model scores well in an experiment. Why is that not the end of AI engineering work?
It still has to become a production system: integrated into an application, deployed for real users, kept reliable and scalable, and owned (monitored, maintained, improved) after deployment.
Scenario: Weeks after release, a deployed model starts producing unexpected outputs. Which responsibility covers finding and fixing this?
System ownership. Monitoring after deployment shows whether the model still operates correctly, and the owner responds with updates or fixes.
Problem-first framing
Clearly defining the system's goal, the needs of users, and the outcomes it should support before selecting technologies or models.
User needs
The goals, expectations, and requirements of the people who interact with or are affected by a system.
Success metrics
Measurable criteria used to determine whether a system achieves its intended goals or performs as expected.
AI system design
Organizing components such as data pipelines, models, and applications so an AI system can function effectively in real-world environments.
Operational environment
The setting in which a system runs and interacts with users, data sources, and other software systems.
Maintainability
The degree to which a system can be updated, repaired, or improved without major disruption to its operation.
Design constraints
Limitations that influence how a system can be built, such as available data, computing resources, time requirements, or organizational policies.
Design trade-offs
Decisions that balance competing priorities, where improving one aspect of a system may reduce another.
Scalability vs reliability vs maintainability: one line each
Scalability: handles more data or users without losing performance. Reliability: consistent, correct results over time. Maintainability: can be updated or repaired without major disruption.
Scenario: An engineer switches to a more complex model. Accuracy improves, but latency and compute cost rise. What concept is this?
A design trade-off. Improving one aspect (accuracy) reduced others (speed and cost).
Scenario: Privacy rules mean the team cannot use certain patient data, so the model has less training data. What concept is this?
A design constraint. Data availability and organizational or privacy policies limit how the system can be built.
Scenario: Before choosing any tools or models, a team meets to define what the system should accomplish, who it serves, and how success will be measured. What is this?
Problem-first framing: goals, user needs, and success metrics are set before technology is selected.
Interdisciplinary systems
Systems that combine knowledge from multiple fields to address complex problems.
Software engineering (in AI engineering)
The discipline focused on designing, building, testing, and maintaining software systems that operate reliably in real-world environments.
Math foundations
The statistical and mathematical methods that support machine learning models and help systems analyze patterns in data.
Ethics in AI
The study and application of principles that guide the responsible design and use of AI systems.
Human-centered design
An approach to system development that prioritizes the needs, understanding, and experiences of the people who interact with the system.
Cross-functional collaboration
Cooperation among professionals from different disciplines who work together to design, develop, and maintain AI systems.
Mental models vs user expectations
Mental models: people's assumptions about how a system works and how it should behave. User expectations: beliefs about how a system should function and what outcomes it should produce.
Sociotechnical systems
Systems that combine technical components, such as software and algorithms, with social elements, including human users, organizational practices, and institutional policies.
In an AI project, what does each role contribute: data scientists, software engineers, designers/product teams?
Data scientists analyze data and develop predictive models. Software engineers build the infrastructure and applications that let models operate in larger systems. Designers and product teams shape how users interact with the system and how information is presented.
Scenario: A hospital deploys an AI tool for prioritizing patients. Staff workflows and organizational policies end up shaping how it is used. What concept does this show?
Sociotechnical systems. AI systems interact with human behavior and organizational context, not just algorithms and data.
Narrow (weak) AI
AI designed for a single or limited task, with no consciousness or understanding. Examples: speech or image recognition, search engines, virtual assistants. This is the AI that exists today.
General (strong) AI
AI that could understand and learn any intellectual task a human can and transfer knowledge between domains. Largely theoretical, not yet realized.
Artificial superintelligence
Hypothetical AI that surpasses human intelligence across all areas. Raises ethical and existential concerns, including loss of control.
Machine learning (ML)
Algorithms that learn from data to identify patterns and make predictions. Includes supervised, unsupervised, and reinforcement learning.
Neural network
A series of algorithms inspired by the human brain that capture relationships in data. Effective at finding patterns in large datasets.
Robotics (as an AI component)
Integrating AI with mechanical or electronic systems so physical machines can perform tasks autonomously or with minimal human help.
Expert system
An AI system that mimics a human expert's decision-making by applying a set of rules to reach conclusions or diagnoses.
Percipio's four ethical considerations in AI
Autonomy in machines, bias and fairness, privacy, and regulation and governance.
Chatbot lab: which competency-2 primitives (variables, functions, data structures) does it use?
Variables: the user's input and the pairs rules. Data structure: a list of pattern-and-response pairs. Functions: a chatbot() function that creates Chat(pairs, reflections) and calls converse(), which loops until the user types quit.
1950 — Turing Test
Alan Turing's paper "Computing Machinery and Intelligence" proposed it: if an evaluator cannot reliably tell the machine from a human, the machine shows intelligent behavior.
1956 — Dartmouth Conference
Where the term "artificial intelligence" was coined. It marks the formal birth of AI as a field of study.
1966 — ELIZA
The first chatbot. It simulated conversation by rephrasing the user's input.
1997 — Deep Blue
IBM's Deep Blue defeated world chess champion Garry Kasparov.
2011 — Watson
IBM's Watson won Jeopardy!, showing advanced natural language processing.
2016 — AlphaGo
DeepMind's AlphaGo defeated world Go champion Lee Sedol.