AWS Certified AI Practitioner Foundational AIF-C01

Transfer learning

A Machine Learning technique where a model that is pre-trained on one task is fine-tuned for a new, related task.

Amazon Aurora

A MySQL and PostgreSQL-compatible relational database designed for high performance and availability, supporting complex applications.

Amazon QuickSight

A business intelligence (BI) tool that you can use to build reports and dashboards, primarily on structured data.

Artificial Intelligence

A field of computer science dedicated to solving cognitive problems commonly associated with human intelligence

Generative AI Security Scoping Matrix

A framework used to identify and manage security risks associated with deploying and using generative AI models.

Amazon Macie

A fully managed data security and privacy service that uses machine learning to discover, classify, and protect sensitive data in AWS.

Amazon Neptune

A fully managed graph database service optimized for storing and querying highly connected datasets, ideal for graph-based applications.

Amazon OpenSearch Service

A fully managed service for search, monitoring, and data analysis, offering real-time search and analytics capabilities for various use cases.

Amazon Transcribe

A fully managed speech-to-text service that automatically converts audio files into accurate, readable text.

International Organization for Standardization (ISO)

A global standard-setting body that provides guidelines and requirements for various industries to ensure quality, safety, and efficiency.

Amazon Personalize

A machine learning service that enables developers to create individualized product recommendations, personalized search results, and targeted marketing promotions based on user behavior and preferences.

Amazon SageMaker Clarify

A machine learning tool that provides insights to improve model fairness and transparency by examining potential bias in datasets and model predictions.

Real-time Inference

A method in Amazon SageMaker for making instant predictions with low latency, suitable for applications requiring immediate responses.

Batch transform

A method in Amazon SageMaker for making predictions on large datasets in bulk, ideal for scenarios where real-time processing isn't required.

Chain-of-Thought

A method in machine learning where models are encouraged to generate reasoning steps before arriving at a final answer, improving decision-making.

Single-Shot

A method where a model processes all necessary information in one go to make a prediction or decision without iterative steps.

Bilingual Evaluation Understudy (BLEU)

A metric for assessing the quality of machine-translated text by comparing the similarity between the machine output and human reference translations.

BERTScore

A metric for evaluating text generation models by comparing similarity at the token level using BERT embeddings.

Perplexity

A metric that you can use to evaluate language models. Perplexity measures the probability of a model to generate a given sequence of words

Mean Squared Error

A metric that you can use to evaluate regression models by measuring the average squared difference between the predicted and actual values

Recall-Oriented Understudy for Gisting Evaluation (ROUGE)

A metric used to evaluate the quality of text summarization by comparing the overlap between generated and reference summaries.

Embedding

A numerical representation of real-world objects that machine learning (ML) and artificial intelligence (AI) systems use to understand complex knowledge domains like humans do

Area Under the ROC Curve (AUC)

A performance metric for classification models, representing the model's ability to distinguish between classes across various thresholds.

Prompt templates

A predefined formats that you can use to standardize inputs and outputs for AI models.

Instruction tuning

A process in machine learning where models are fine-tuned to follow explicit instructions, improving their performance on specific tasks.

Amazon Inspector

A security assessment service that automatically scans AWS workloads for vulnerabilities and deviations from best practices, helping ensure compliance and security.

Prompt Injection

A security vulnerability where malicious input is embedded in prompts to manipulate the output of language models.

Amazon Augmented AI (Amazon A2I)

A service that makes it easy to incorporate human review into machine learning predictions, ensuring high-quality results.

AWS PrivateLink

A service that provides private connectivity between VPCs and AWS services, ensuring secure and scalable access to critical services.

Intelligent Document Processing (IDP)

A software tool that extracts and categorizes information from unstructured or structured data, generates summaries, and delivers actionable insights.

System and Organization Controls (SOC)

A suite of reports that provide information about the controls at a service organization, relevant to security, availability, processing integrity, confidentiality, or privacy.

Classification

A supervised learning technique that assigns labels or categories to new, previously unseen data examples using a learned model.

Regression

A supervised learning technique that predicts continuous or numerical values given one or more input variables.

Asynchronous Inference

A technique in Amazon SageMaker that allows models to process inference requests in a non-blocking manner, handling large or delayed responses efficiently.

Instruction-based fine-tuning

A technique that uses labeled examples to improve model performance on a specific task.

Zero-Shot

A technique where a model is able to perform tasks it hasn't been explicitly trained for, leveraging knowledge from related tasks.

Conversational AI

A technology that makes software capable of understanding and responding to voice-based or text-based human conversations

Amazon Polly

A text-to-speech service that converts written content into lifelike speech, enabling the creation of voice-enabled applications.

Generative AI

A type of AI that can create new content and ideas, including conversations, stories, images, videos, and music

Amazon SageMaker Feature Store

Amazon SageMaker Feature Store is a fully managed, purpose-built repository for storing, sharing, and managing machine learning (ML) features.

Managed API Service

An API service hosted and maintained by a third-party provider, offering scalability, security, and reduced operational overhead.

Self-Hosted API

An API that is deployed and managed by the organization itself, giving full control over the infrastructure and configurations.

Data privacy regulations

An important component of any business to ensure customer and personal data remains private.

Dimensionality reduction

An unsupervised learning strategy that minimizes the number of features or dimensions in a dataset while retaining the most relevant information or patterns.

Amazon SageMaker Autopilot

Automates the creation, training, and tuning of ML models. You can use SageMaker Autopilot to develop high-quality models with minimal effort. However, automating the creation, training, and tuning does not help you create a trusted training dataset that incorporates human input and feedback.

Amazon SageMaker Model Monitor

Automatically monitors machine learning (ML) models in production and alerts you when quality concerns arise. Model Monitor applies criteria to detect drift in your models and warns you when it occurs.

AWS AI Service Cards

Describe FMs that are developed by AWS AI services. Service cards can include sample performance data for commonly used benchmarking datasets.

Amazon SageMaker Model Cards

Documents that provide key details about machine learning models, including performance metrics, intended use, and compliance information.

Negative prompts

In generative models, these prompts specify what content should be excluded from the generated output.

Low bias

Indicates that the model is not making erroneous assumptions about the training data.

High variance

Indicates that the model is paying attention to noise in the training data and is overfitting.

Feature selection

Involves the selection of data attributes or variables during the development of a predictive model.

Traditional machine learning models

It accomplish tasks using the data you provide. They can make predictions based on ranking, sentiment analysis, image categorization, and other factors. However, each model is limited to performing a single task.

Adaptability

It can adapt to a variety of activities and domains by learning from data and producing material that is suited to specific situations or needs. Because of its flexibility, generative AI can be applied to a wide number of sectors.

Creativity and exploration

It can develop new ideas, designs, or solutions by combining and recombining pieces in unusual ways. This can encourage creativity and the discovery of new possibilities.

Responsiveness

It can generate content in real time, resulting in faster reaction times and more dynamic interactions. This is especially beneficial for chatbots, virtual assistants, and other interactive applications that demand instant feedback.

Data efficiency

It can learn from relatively little quantities of data and produce new samples that are consistent with the training data. This can be useful when data is limited or difficult to collect.

Simplicity

It can make hard tasks easier by automating content generation processes. For example, AI language models may generate human-like text, reducing the time and effort necessary for content development.

Personalization

It can may develop personalized content based on individual preferences or attributes, hence improving user experiences and engagement.

Computer Vision

It enables machines to recognize people, places, and things in photographs with accuracy equal to or greater than that of humans, while also operating at considerably higher speeds and efficiency.

Natural Language Processing (NLP)

It is a field within artificial intelligence focused on the interaction between computers and human languages.

Amazon Bedrock

It is a fully managed solution that makes high-performing foundation models (FMs) from top AI startups and Amazon available to you through a common API.

Amazon Q Business

It is a generative AI-powered assistant that can answer queries, generate content, provide summaries, and complete tasks based on the data in your organization.

Amazon Comprehend

It is a managed service that uses natural language processing (NLP) tasks to extract key phrases, sentiments, or key findings in text.

Retrieval Augmented Generation

It is a method that can optimize the output of a large language model (LLM) by referencing a knowledge base with company-specific or industry-specific data.

Domain adaptation fine-tuning

It is a method that you can use to customize a pre-trained FM by fine-tuning the model on a specific task or domain-specific information.

AWS Key Management Service

It is a service that you can use to store and manage cryptographic keys. You can create and control the keys that protect data.

Deep Learning

It is a technique in artificial intelligence (AI) that teaches computers to process data in a manner inspired by the human brain.

Reinforcement Learning from Human Feedback (RLHF)

It is a technique that incorporates human feedback to help ML models more efficiently and accurately make predictions and maximize rewards.

Few-shot Prompt Engineering

It is a technique that you can use to guide a model to generalize based on a few examples. The model uses examples to generalize and make more accurate predictions without the need to re-train or fine-tune a model.

AWS Trusted Advisor

It is a tool that inspects your AWS environment and provides recommendations to improve security, optimize cost, or improve system availability.

AWS Secrets Manager

It is a tool that you can use to manage and maintain credentials. You can use Secrets Manager to improve security by rotating the keys and tokens that you use in an AWS environment.

PartyRock

It is an Amazon Bedrock Playground that allows users to build generative AI apps easily and intuitively. The platform provides a fun, hands-on environment where users can create various AI-driven applications in just a few steps.

Neural Networks

It is an artificial intelligence technique that teaches computers to interpret data in a manner inspired by the human brain.

Fine-tuning

It is the process of training a previously trained model on a fresh dataset rather than starting from scratch. This approach, also known as transfer learning, can generate reliable models from smaller datasets and require less training time.

Machine Learning

It is type of AI for understanding and building methods that make it possible for machines to learn.

Amazon Lex

It is used to create chatbots based on intents and entities. The chatbots could potentially provide answers based on internal documentation.

Amazon SageMaker

It provides a complete machine learning lifecycle, including data preparation, model building, training, tuning, and deployment.

Responsible AI

It refers to the procedures and principles that ensure AI systems are transparent and trustworthy while minimizing potential risks and bad effects.

Amazon SageMaker Ground Truth

It uses human feedback to create labeled datasets. By incorporating human-verified labels and human oversight, SageMaker Ground Truth helps align the AI's decision-making process more closely with real-world contexts.

Recall

Measures how well an algorithm correctly predicts all the true positives (TP) in a dataset. You can use recall for clas

Reinforcement learning

One continuously improves their model by analyzing feedback from prior versions. In reinforcement learning, an agent learns by trial and error as it interacts with its surroundings.

Amazon SageMaker Pipelines

Provides a framework to orchestrate the ML model development workflow. You can use SageMaker Pipelines to create consistent, repeatable, and scalable ML operations.

AWS Config

Provides an overview of your AWS resource configurations. You can use AWS Config to identify how resources were configured in the past.

Model Latent Space

The abstract space where machine learning models map input data into feature representations used to generate outputs.

Unsupervised learning

The algorithm tries to discover hidden patterns or structures within the data without any prior information or guidance.

Nondeterminism

The concept where outcomes are not entirely predictable, often due to randomness or external factors influencing the process.

Mean Absolute Percenage Error (MAPE)

The mean of the absolute differences between the actual values and the predicted values, divided by the actual values.

Exploratory Data Analysis (EDA)

The process of analyzing data sets to summarize their main characteristics, often using visualizations to uncover patterns, trends, and relationships.

Fraud Detection

The process of identifying and preventing fraudulent activities using techniques like machine learning, anomaly detection, and pattern recognition.

Hyperparameter Tuning

The process of optimizing the parameters that control the learning process in machine learning models to improve performance.

Prompt Engineering

The process where you guide generative artificial intelligence (generative AI) solutions to generate desired outputs.

Model validation

The step where you determine if model performance and accuracy is appropriate for your use case before you deploy the model.

Factors to consider when selecting a generative AI model.

There are 5 factors to consider when selecting a model; Performance requirements, Constraints, Capabilities, Compliance, Cost

Multimodal Models

These models are designed to process inputs from various sources, including text, images, audio, and video.

Supervised learning

This is a prominent type of machine learning due to its vast range of applications. It's termed supervised learning because there must be a supervisor.

Clustering

This method divides data into clusters based on similar traits or distances between data points in order to better understand the characteristics of a particular cluster.

Challenges of generative AI

Toxicity, Hallucinations, Intellectual Property, Plagiarism and Cheating, lastly Disruptions of Nature Work.

Scalability

When trained, generative AI models may produce a vast amount of information quickly. This makes the models suited for situations requiring large-scale content production.

Amazon SageMaker JumpStart

With JumpStart, you can rapidly evaluate, compare, and pick FMs for activities like article summarizing and picture production based on predefined quality and responsibility measures.