[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"cheat-sheet---en":3,"domain-info---en":3,"topic-info----en":3,"prev-aws-certified-ai-practitioner-ai-ml-fundamentals-ai-ml-core-concepts-what-is-ai-and-ml-en":3,"lesson-aws-certified-ai-practitioner-ai-ml-fundamentals-ai-ml-core-concepts-what-is-ai-and-ml-en":4,"next-aws-certified-ai-practitioner-ai-ml-fundamentals-ai-ml-core-concepts-what-is-ai-and-ml-en":255},null,{"locked":5,"reason":3,"meta":6,"item":17},false,{"title":7,"description":8,"isFree":9,"estimatedMinutes":10,"difficulty":11,"learningObjectives":12},"What Is AI and ML?","Define artificial intelligence and machine learning, see how they differ from ordinary programming, and learn how AI, ML, deep learning, and generative AI nest inside each other.",true,16,"beginner",[13,14,15,16],"Define artificial intelligence and machine learning and explain how each differs from rule-based programming","Explain how AI, ML, deep learning, and generative AI nest as subsets of one another","Identify the core vocabulary of ML: model, algorithm, training, and inference","Decide when a problem calls for machine learning instead of hand-written rules",{"id":18,"title":7,"body":19,"description":8,"difficulty":11,"estimatedMinutes":10,"extension":195,"infographics":196,"isFree":9,"learningObjectives":207,"meta":208,"navigation":9,"path":209,"quiz":210,"seo":252,"stem":253,"__hash__":254},"courses/courses/aws-certified-ai-practitioner/en/domains/01-ai-ml-fundamentals/01-ai-ml-core-concepts/01-what-is-ai-and-ml.md",{"type":20,"value":21,"toc":184},"minimark",[22,26,29,34,37,40,44,47,50,53,57,60,65,94,97,100,104,107,114,121,127,133,136,147,151,154,157,160,164,167,181],[23,24,25],"p",{},"A bank wants to flag fraudulent card transactions. The old way was to write rules: block any purchase over $5,000, block purchases from two countries within an hour, and so on. Fraudsters learn the rules and slip under them, so the team keeps adding more rules until nobody understands the system. The patterns that actually signal fraud are subtle, shifting, and buried in millions of transactions. No human can write them all down.",[23,27,28],{},"This is the gap machine learning fills. Instead of writing the rules yourself, you show a system many past transactions labeled \"fraud\" or \"legitimate,\" and it works out the pattern on its own. That shift, from writing rules to learning them from data, is the whole idea behind the field this exam is built on. This lesson gives you the vocabulary the rest of the course depends on.",[30,31,33],"h2",{"id":32},"what-artificial-intelligence-is","What artificial intelligence is",[23,35,36],{},"Artificial intelligence is the broad umbrella term for any technique that makes a machine perform tasks that normally need human intelligence. AWS describes AI as \"an umbrella term for different strategies and techniques you can use to make machines more humanlike.\" That includes recognizing a face in a photo, understanding a spoken question, recommending a movie, or driving a car.",[23,38,39],{},"AI is deliberately broad. It covers rule-based systems (hand-written logic), search algorithms, and machine learning, among others. The key point for the exam: not everything called AI learns from data. A thermostat that follows a fixed schedule is a simple rule-based system. It is automation, and you can argue it is AI, but it is not machine learning, because nothing about it improves from experience.",[30,41,43],{"id":42},"what-machine-learning-is","What machine learning is",[23,45,46],{},"Machine learning is the branch of AI that learns patterns from data instead of following rules a developer wrote by hand. AWS defines it as \"the science of developing algorithms and statistical models that computer systems use to perform complex tasks without explicit instructions,\" relying on \"patterns and inference instead.\"",[23,48,49],{},"Hold onto that phrase: without explicit instructions. In traditional programming, you write the logic and the computer follows it. In machine learning, you provide data and desired outcomes, and the system produces the logic. Consider spam filtering. The rule-based approach is a list like \"block any email containing 'free money.'\" The ML approach is to feed the system thousands of emails already marked spam or not spam, and let it learn which combinations of words, senders, and patterns predict spam. When spam evolves, you retrain on new examples instead of rewriting rules.",[23,51,52],{},"That is the trade the whole field is built on: you give up direct control over the logic, and in return you get a system that can handle patterns too complex or too changeable to write down.",[30,54,56],{"id":55},"how-ai-ml-deep-learning-and-generative-ai-fit-together","How AI, ML, deep learning, and generative AI fit together",[23,58,59],{},"These four terms are not four separate things. They nest inside each other, largest to smallest, and the exam will test whether you know the order.",[61,62],"infographic",{"alt":63,"slug":64},"Concentric rings showing Generative AI inside Deep Learning inside Machine Learning inside Artificial Intelligence, each labeled with a one-line differentiator.","ai-ml-dl-genai-nesting",[66,67,68,76,82,88],"ul",{},[69,70,71,75],"li",{},[72,73,74],"strong",{},"Artificial intelligence"," is the outermost ring. Any technique that makes machines act intelligently belongs here.",[69,77,78,81],{},[72,79,80],{},"Machine learning"," sits inside AI. It is the subset that learns from data. All machine learning is AI, but not all AI is machine learning.",[69,83,84,87],{},[72,85,86],{},"Deep learning"," sits inside ML. It is the subset that uses neural networks with many layers (the next lesson covers this in detail). All deep learning is machine learning.",[69,89,90,93],{},[72,91,92],{},"Generative AI"," sits inside deep learning. It is the subset that creates new content: text, images, audio, and code. All generative AI is deep learning.",[23,95,96],{},"Read the rings from the inside out and each one \"is a kind of\" the ring around it. Generative AI is a kind of deep learning, which is a kind of machine learning, which is a kind of AI. Agentic AI, which you will meet in a later lesson, builds on generative AI by letting models plan and take actions, so it lives near the center too.",[23,98,99],{},"The common mistake is treating these as rivals (\"should I use AI or ML?\"). They are not alternatives. When you use generative AI, you are using all four layers at once.",[30,101,103],{"id":102},"the-core-vocabulary-model-algorithm-training-inference","The core vocabulary: model, algorithm, training, inference",[23,105,106],{},"Four words appear on nearly every page of this course. Learn them once here.",[23,108,109,110,113],{},"An ",[72,111,112],{},"algorithm"," is the procedure that learns from data. It is the recipe, such as \"linear regression\" or \"a neural network.\" You pick an algorithm before you have a model.",[23,115,116,117,120],{},"A ",[72,118,119],{},"model"," is what you get after an algorithm learns from your data. It is the trained artifact that holds the learned patterns and turns new inputs into predictions. The algorithm is the recipe; the model is the cake you baked with your specific ingredients.",[23,122,123,126],{},[72,124,125],{},"Training"," is the phase where the algorithm processes your data and produces the model. Training is usually the compute-heavy, done-ahead-of-time part. You train a fraud model overnight on last year's transactions.",[23,128,129,132],{},[72,130,131],{},"Inference"," is the phase where the finished model makes predictions on new, unseen inputs. When a live transaction arrives and the model scores it as fraud or not, that is inference. A whole later lesson covers the different ways to run inference, because that choice drives cost and latency in production.",[23,134,135],{},"Keep training and inference separate in your mind. Training builds the model; inference uses it. Many exam questions hinge on knowing which phase a described activity belongs to.",[23,137,138,139,142,143,146],{},"Two more terms round out the basic glossary, and both get their own depth later in the course. ",[72,140,141],{},"Bias"," is a systematic error that makes a model unfair or inaccurate for certain groups. ",[72,144,145],{},"Fairness"," is the goal of a model treating different groups equitably. Both belong to the responsible-AI domain, so just recognize them for now.",[30,148,150],{"id":149},"when-machine-learning-is-the-right-tool-and-when-it-is-not","When machine learning is the right tool (and when it is not)",[23,152,153],{},"Machine learning is powerful, but it is not the answer to everything, and the exam rewards knowing the difference. Reach for ML when three things are true: the pattern is too complex or changeable to write as rules, you have enough relevant data to learn from, and an approximate prediction is acceptable rather than a guaranteed exact answer.",[23,155,156],{},"Fraud detection, product recommendations, and demand forecasting all fit. The patterns are messy, the data is plentiful, and a good prediction is genuinely useful even when it is occasionally wrong.",[23,158,159],{},"Now the flip side. If you need to calculate sales tax, do not use machine learning. Tax is a fixed rule, the answer must be exact, and a model that is \"97% accurate\" on tax is a broken system. When the logic is simple and known, and the output must be provably correct every time, write the rule. ML trades exactness for the ability to handle complexity, and that trade only makes sense when the complexity is real.",[30,161,163],{"id":162},"exam-tips","Exam tips",[23,165,166],{},"The exam probes the boundaries between these terms far more than the definitions themselves.",[66,168,169,172,175,178],{},[69,170,171],{},"When a question lists AI, ML, deep learning, and generative AI and asks about their relationship, the answer is almost always the nesting order (each is a subset of the one before). Reject any option that makes them separate or reverses the containment.",[69,173,174],{},"Watch for the keyword \"learns from data\" or \"without explicit rules.\" That points at machine learning, not plain AI or rule-based automation.",[69,176,177],{},"If a scenario needs a guaranteed exact output (tax, a legal calculation, a fixed business rule), the exam wants you to say ML is not appropriate. The tell is a phrase like \"a specific outcome is needed\" rather than a prediction.",[69,179,180],{},"Do not confuse \"algorithm\" with \"model.\" The algorithm is chosen before training; the model is produced by training.",[23,182,183],{},"The one idea to carry forward: machine learning replaces hand-written rules with patterns learned from data, and it sits inside the wider field of AI while containing deep learning and generative AI within itself. Every later lesson builds on this nesting, starting with the neural networks that make deep learning possible.",{"title":185,"searchDepth":186,"depth":186,"links":187},"",3,[188,190,191,192,193,194],{"id":32,"depth":189,"text":33},2,{"id":42,"depth":189,"text":43},{"id":55,"depth":189,"text":56},{"id":102,"depth":189,"text":103},{"id":149,"depth":189,"text":150},{"id":162,"depth":189,"text":163},"md",[197],{"slug":64,"concept":198,"style":199,"aspectRatio":200,"labels":201},"A single set of concentric rings on a dark canvas showing containment, largest to smallest: Artificial Intelligence as the outer ring, Machine Learning inside it, Deep Learning inside that, and Generative AI as the innermost ring. Each ring is labeled with its name and a one-line differentiator. A short footer strip states the containment rule. The emphasis is that each inner ring is fully inside the outer one, so the reader reads the nesting as 'is a kind of' from inside out.","diagram","16:9",[202,203,204,205,206],"Artificial Intelligence: any technique that makes machines perform tasks that need human-like intelligence","Machine Learning: systems that learn patterns from data instead of following hand-written rules","Deep Learning: ML built on neural networks with many layers","Generative AI: deep learning that creates new text, images, audio, and code","Each ring sits fully inside the next: all generative AI is deep learning, all deep learning is ML, all ML is AI",[13,14,15,16],{},"/courses/aws-certified-ai-practitioner/en/domains/01-ai-ml-fundamentals/01-ai-ml-core-concepts/01-what-is-ai-and-ml",{"passingScore":211,"questions":212},70,[213,222,230,238,244],{"question":214,"type":215,"options":216,"correctAnswer":218,"explanation":221},"What is the core difference between machine learning and traditional rule-based programming?","single",[217,218,219,220],"Machine learning runs in the cloud, while rule-based programming runs on local servers","Machine learning learns patterns from data, while rule-based programming follows rules a developer wrote by hand","Machine learning is always more accurate than rule-based programming","Machine learning does not need any data to work","In rule-based programming, a developer writes the logic explicitly (if this, then that). In machine learning, you show the system labeled or raw data and it derives the pattern itself. The first answer confuses where code runs with how it is built, and machine learning is not automatically more accurate. Data is central to ML, not optional.",{"question":223,"type":215,"options":224,"correctAnswer":227,"explanation":229},"Which statement correctly describes how AI, ML, and deep learning relate?",[225,226,227,228],"They are three separate fields with no overlap","Deep learning contains machine learning, which contains AI","Machine learning is a subset of AI, and deep learning is a subset of machine learning","AI and machine learning are the same thing with different names","The relationship nests from the outside in: AI is the broad umbrella, ML is one branch of AI that learns from data, and deep learning is the branch of ML that uses many-layered neural networks. The second option reverses the containment, and AI includes techniques (like rule-based systems) that are not machine learning.",{"question":231,"type":215,"options":232,"correctAnswer":234,"explanation":237},"In machine learning terms, what is a 'model'?",[233,234,235,236],"The raw dataset used to teach the system","The learned artifact, produced by training, that turns new inputs into predictions","The physical server that runs predictions","The person who labels the training data","A model is the output of the training process: it captures the patterns the algorithm found in the data and uses them to make predictions on new inputs. The dataset is the input to training, not the model itself, and the server and labeler are separate parts of the workflow.",{"question":239,"type":215,"options":240,"correctAnswer":241,"explanation":243},"Training and inference are two distinct phases in the machine learning lifecycle.",[241,242],"True","False","Training is the phase where the model learns patterns from data, which is usually compute-heavy and done ahead of time. Inference is the phase where the finished model makes predictions on new inputs in production. Keeping these two phases separate is a recurring theme across the exam.",{"question":245,"type":215,"options":246,"correctAnswer":248,"explanation":251},"A team needs to detect fraudulent transactions where fraud patterns keep shifting and are hard to describe with fixed rules. Why is machine learning a good fit here?",[247,248,249,250],"Because rules are cheaper to maintain than a model","Because the patterns are complex and changing, so a system that learns from data adapts better than fixed rules","Because machine learning never produces false positives","Because fraud detection does not require any historical data","Fraud is a classic ML use case: the signal is buried in complex, evolving patterns that would take endless hand-written rules to capture. A model learns those patterns from historical transactions and can be retrained as fraud shifts. ML does produce false positives, and it depends heavily on historical data to learn from.",{"title":7,"description":8},"courses/aws-certified-ai-practitioner/en/domains/01-ai-ml-fundamentals/01-ai-ml-core-concepts/01-what-is-ai-and-ml","oHhTReMku5yZ30CjPYL2hHXHIx1E_2WDfr0ItyxptSE",{"locked":9,"reason":256,"meta":257,"item":3},"paywall",{"title":258,"description":259,"isFree":5,"estimatedMinutes":10,"difficulty":11,"learningObjectives":260},"Neural Networks and Deep Learning","Understand what a neural network is, what makes deep learning 'deep,' and how deep learning differs from traditional machine learning on feature engineering, data needs, and interpretability.",[261,262,263,264],"Explain what a neural network is and how neurons, layers, and weights work together","Describe what makes deep learning 'deep' and why depth helps with complex data","Compare deep learning with traditional machine learning on feature engineering, data, and interpretability","Identify tasks where deep learning is the right approach"]