Prepare for OCI AI Foundations

14. What is the primary role of Graphics Processing Units (GPUs) in modern computing? Handling network communications Providing power backup and uninterruptible power supply Accelerating graphics rendering and parallel processing Managing storage devices

Accelerating graphics rendering and parallel processing GPUs are specialized hardware designed primarily for accelerating tasks related to graphics rendering and parallel processing.

15. Which AI domain can be used in the detection of fraudulent transactions? Natural Language Processing Image recognition Anomaly Detection Learn by reward

Anomaly Detection Anomaly detection allows AI systems to continuously analyze and identify potentially fraudulent transactions, providing a valuable tool for financial institutions and businesses to protect against fraud.

6. What is the primary capability of the OCI Language service? Extract text in images Perform speech analysis Change the font specification used in text Process unstructured data and extract insights

Process unstructured data and extract insights OCI Language service is primarily used to process unstructured data and extract insights from text.

What is "in-context learning" in the context of large language models (LLMs)? Training a model on a diverse range of tasks. Modifying the behavior of a pre-trained LLM permanently. Teaching the model through zero-shot learning. Providing a few examples of a target task via the input prompt.

Providing a few examples of a target task via the input prompt. Explanation: In-context learning refers to the capability of large language models (LLMS) to learn and perform new tasks without further training or fine-tuning. Instead of modifying the model permanently, users can guide the model's behavior by providing a few examples of the target task through the input prompt. This is particularly useful when direct access to the model is limits, such as when using it through an API or user interface.

5. What is the primary problem associated with vanishing gradients in Recurrent Neural Networks (RNNs)? RNNs get extremely small gradients during backpropagation. RNNs struggle to handle long sequences of data. RNNs become computationally expensive. RNNs tend to overfit the training data.

RNNs get extremely small gradients during backpropagation. The vanishing gradient problem in RNNs occurs when the gradients (used for updating weights during training) become very small as they are backpropagated through time. This can result in the network having difficulty learning long-range dependencies in sequential data.

8. Which deep learning architecture is well-suited for processing sequential data such as sentences? Recurrent Neural Network (RNN) Variational AutoEncoder (VAE) Generative Adversarial Network (GAN) Convolutional Neural Network (CNN)

Recurrent Neural Network (RNN) Recurrent Neural Networks (RNNs) are designed to handle sequences of data, making them suitable for tasks involving sequential data such as language processing and audio analysis. They can retain contextual information over time, which is beneficial for understanding sequences.

What is the main distinction between classification and regression in supervised Machine Learning? Classification predicts continuous values; regression assigns data points to categories. Classification assigns data points to categories; regression predicts continuous values. Classification and regression both predict continuous values. Classification and regression both assign data points to categories.

Classification assigns data points to categories; regression predicts continuous values. Explanation: The key difference between classification and regression lies in the nature of the target variable. Classification deals with categorical outcomes and assigns data points to specific categories or classes, while regression deals with continuous numeric outcomes and predicts values within a range.

1. How do hidden layers in neural networks help with character recognition? By increasing the network's processing speed By directly mapping input characters to output predictions By improving the interpretability of the model's decisions By enabling the network to learn complex features such as edges and shapes

Correct Option: By enabling the network to learn complex features like edges and shapes Hidden layers in neural networks are crucial for character recognition because they enable the network to learn and extract complex features and patterns, such as edges, shapes, and curves, which are essential for recognizing characters.

7. When preparing data for a Machine Learning model, what are typically considered as input features? Attributes that provide information for making predictions Data that is irrelevant to the problem The target variable you want to predict Data collected after the model has made predictions

Attributes that provide information for making predictions Input features, also known as independent variables, are the attributes or characteristics in your dataset that are used to make predictions. These features are used by a Machine Learning model to learn patterns and relationships in the data and to generate predictions or classifications for the target variable.

3. What is OCI Document Understanding primarily used for? Managing cloud infrastructure resources Generating computer-generated art Analyzing astronomical data Automating document processing tasks

Automating document processing tasks OCI Document Understanding is a service in Oracle Cloud Infrastructure designed for automating document processing tasks, such as extracting information from documents, automating data entry, and streamlining document-based workflows. It's used for tasks such as invoice processing, form recognition, and document classification.

1. Which statement accurately describes generative AI? Focuses on making accurate predictions based on training data Limits functions to generating only text-based outputs Creates new content without making predictions Exclusively trains to predict future data patterns

Correct Option: Creates new content without making predictions Generative AI is focused on creating new content or data rather than making predictions based on existing training data. It involves generating novel and meaningful outputs such as images, text, music, or other forms of creative content.

5. What is the advantage of using OCI Superclusters for AI workloads? Provide a cost-effective solution for simple AI tasks Are ideal for tasks such as text-to-speech conversion Offer seamless integration with social media platforms Deliver exceptional performance and scalability for complex AI tasks

Correct Option: Deliver exceptional performance and scalability for complex AI tasks OCI AI Superclusters are specifically designed to handle demanding AI workloads that require significant computational power and scalability. They are optimized to provide high performance for complex tasks such as training large Machine Learning models, deep learning, and other compute-intensive AI tasks.

3. Which OCI AI service is used to extract tabular content from documents? Image Analysis Object Detection Document Understanding Speech

Correct Option: Document Understanding Document Understanding is an AI service offered by OCI that specializes in processing and extracting structured content from unstructured documents. This includes extracting tabular data, as well as text and other structured information, from documents such as PDFs, images, and scanned documents.

4. Sequence models are used to solve problems involving sequentially ordered data points or events. Which is NOT the best use case for sequence models? Image classification and object recognition Speech recognition and language translation Time series analysis and forecasting Natural language processing tasks such as sentiment analysis

Correct Option: Image classification and object recognition Sequence models are indeed well-suited for tasks involving sequentially ordered data points or events, such as time series analysis, natural language processing, speech recognition, and language translation. However, for image classification and object recognition, traditional machine learning models and convolutional neural networks (CNNs) are more commonly used.

1. Which OCI Data Science feature enables you to define and run repeatable Machine Learning tasks on fully managed infrastructure? Conda Environments Model Catalog Jobs Model Detection

Correct Option: Jobs Jobs in OCI Data Science allows you to define and run repeatable Machine Learning tasks and workflows. You can create and execute specific operations, such as data preprocessing, model training, model evaluation, and more, using Jobs. They provide a structured way to automate and manage individual tasks within a data science project.

2. Which sequence model can maintain relevant information over long sequences? Convolutional Neural Networks Long Short-Term Memory Neural Networks Recurrent Neural Networks Feed Forward Neural Networks

Correct Option: Long Short-Term Memory Neural Networks Long Short-Term Memory (LSTM) Neural Networks are the sequence model of choice when it comes to handling and maintaining relevant information over long sequences, making them particularly well-suited for tasks such as language modeling, machine translation, and speech recognition.

2. Which language is NOT supported by the OCI Speech service? Spanish Portuguese English Mandarin

Correct Option: Mandarin OCI Speech Supports English, Spanish, and Portuguese.

2. Which OCI Data Science feature allows you to use catalogued models as HTTP endpoints on fully managed infrastructure? Jobs Model Catalog Conda Environments Model Deployments

Correct Option: Model Deployments Model Deployments in OCI Data Science enable you to deploy your Machine Learning models as HTTP endpoints, making them accessible for real-time predictions and inferences. You can easily deploy, manage, and scale these models on fully managed infrastructure. Model Deployments are a key component for operationalizing your Machine Learning models and integrating them into your applications or services.

2. Which aspect of Large Language Models significantly impacts their capabilities, performance, and resource requirements? Number of training iterations performed during model training Total number of GPUs used for model training Model size and parameters, including the number of tokens and weights Complexity of the programming languages used for model development

Correct Option: Model size and parameters, including the number of tokens and weights. The size and complexity of a language model, including the number of parameters (weights) and tokens have a profound impact on its capabilities and performance. Larger models with more parameters tend to have a better understanding of language and can generate more coherent and contextually relevant text. Larger models, however, require substantial computational resources, including GPUs and memory, for both training and inference.

5. Which essential component of Artificial Neural Network performs weighted summation and applies activation function on input data to produce an output? Iterator Classifier Bias Neuron

Correct Option: Neuron A neuron in an Artificial Neural Network is the fundamental building block responsible for performing weighted summation and applying an activation function to input data to produce an output.

4. What is the primary value proposition of Machine Learning in Oracle Database? Focuses on transferring data and providing flexible architecture to enhance database performance and scalability Offers a complex pricing structure with additional costs for Machine Learning features Eliminates data movement, empowers users with Machine Learning, and offers a simpler architecture Provides algorithms specifically redesigned for data movement in hybrid environments

Correct Option: OML eliminates data movement, empowers users with machine learning, and offers a simpler architecture. Oracle Database's Machine Learning capabilities are designed to eliminate the need to move data out of the database for Machine Learning tasks. This is a significant advantage because it reduces data latency, enhances security, and simplifies the overall architecture of data-driven applications. By providing in-database Machine Learning, Oracle empowers users to perform Machine Learning tasks directly within the database, leveraging its computational power and efficiency.

4. Which capability of OCI Vision service uses a bounding box inside an image? Image Classification Image Recognition Object Repair Object Detection

Correct Option: Object Detection Object Detection is a computer vision task that involves identifying and locating objects within an image. It not only recognizes the presence of objects but also provides information about their location in the image through the use of bounding boxes. Each bounding box represents the position and size of a detected object within the image.

5. What type of clustering algorithm is used to cluster the data points into nonoverlapping clusters? Density based Partition based Distribution based Weight based

Correct Option: Partition based Partition-based clustering algorithms (option c) are used to divide data points into nonoverlapping clusters, where each data point belongs to exactly one cluster. The most well-known example of a partition-based clustering algorithm is K-Means. K-Means iteratively assigns data points to the nearest cluster center and then recomputes cluster centers until convergence, resulting in distinct, nonoverlapping clusters.

1. Which application does NOT require a Machine Learning solution? Password Validation Detecting spam emails Stock Price Predictions Customer Segmentation

Correct Option: Password Validation Password Validation (option a) typically does not require a Machine Learning solution. Password validation is a straightforward process that involves checking whether a user-entered password matches the stored password on a server. This can be achieved through standard cryptographic techniques and rules without the need for Machine Learning.

5. What is "in-context learning" in the context of large language models (LLMs)? Providing a few examples of a target task via the input prompt Training a model on a diverse range of tasks Teaching the model through zero-shot learning Modifying the behavior of a pretrained LLM permanently

Correct Option: Providing a few examples of a target task via the input prompt. In-context learning refers to the capability of generative large language models (LLMs) to learn and perform new tasks without further training or fine-tuning. Instead of modifying the model permanently, users can guide the model's behavior by providing a few examples of the target task through the input prompt. This is particularly useful when direct access to the model is limited, such as when using it through an API or user interface.

3. Which neural network has a feedback loop and is designed to handle sequential data? Multi-Layer Perceptron Neural Networks Feed Forward Neural Networks Convolution Neural Networks Recurrent Neural Networks

Correct Option: Recurrent Neural Networks Recurrent Neural Networks (RNNs) are a type of neural network architecture that includes feedback connections. These feedback connections allow RNNs to process sequential data such as time series, natural language, speech, and more.

4. What type of Machine Learning algorithm is used when we want to predict the resale price on a residential property? Regression Multiclass Classification Binary Classification Anomaly Detection

Correct Option: Regression Regression (option a) is the type of Machine Learning algorithm used when we want to predict continuous numerical values such as the resale price of a residential property. In regression tasks, the goal is to learn a mapping between input features (for example, square footage, number of bedrooms, and location) and a continuous target variable (for example, price). Linear Regression is most employed for this purpose.

2. Which type of Machine Learning is used in autonomous car driving? Reinforcement Learning Natural Language Processing Unsupervised Learning Supervised Learning

Correct Option: Reinforcement Learning Reinforcement Learning (RL) is a branch of Machine Learning where an agent learns to perform actions in an environment to maximize a cumulative reward. In the context of autonomous car driving, the car is the agent, the road and its surroundings form the environment, and the reward might be related to safe and efficient navigation, obeying traffic rules, and reaching the desired destination.

2. Which type of Machine Learning algorithm learns from outcomes to make decisions? Supervised Learning Natural Language Processing Reinforcement Learning Unsupervised Learning

Correct Option: Reinforcement learning Reinforcement Learning (option c) is a type of Machine Learning algorithm that learns from outcomes to make decisions. In Reinforcement Learning, an agent interacts with an environment and takes actions to maximize cumulative rewards.

3. Which is NOT an example of vision or image-related AI task? Identify boundaries in an image Repair damaged images Identify objects in images Classify images

Correct Option: Repair damaged images While image restoration and repair do involve working with images, it's not a typical vision task in the sense of identifying objects, classifying images, or facial recognition. While AI can certainly be used for image restoration, it's not one of the core tasks directly related to vision or image-related AI.

3. Which type of function is used in Logistic Regression to predict a loan defaulter? Identity function Sigmoidal function Gaussian function Step function

Correct Option: Sigmoidal function to predict the probability of binary outcome Logistic Regression is a binary classification algorithm commonly used in Machine Learning to predict binary outcomes, such as whether a loan will be defaulted or not. The key idea behind Logistic Regression is to model the probability of an event occurring as a function of input features. The output of this model is transformed using the sigmoidal (also known as logistic) function. The sigmoidal function, often represented as the sigmoid function, has an S-shaped curve that maps any input value to an output value between 0 and 1.

4. Which task is an example of a speech-related AI task? Random number generation Speech-to-text conversion Language translation Music composition

Correct Option: Speech-to-text Conversion This task involves converting spoken language or speech into written text. It's commonly referred to as automatic speech recognition (ASR) or speech-to-text conversion. Speech-related AI algorithms process audio recordings of spoken words and transcribe them into textual form.

3. Fine-tuning is unnecessary for Large Language Models (LLMs) if your application does not involve which specific aspect? Efficiency & resource utilization Bias mitigation Task-specific adaptation Domain vocabulary

Correct Option: Task-specific adaptation Fine-tuning of Large Language Models (LLMs) is primarily performed to adapt the model to specific tasks or domains. If your application doesn't require task-specific adaptation, then fine-tuning may not be necessary. Fine-tuning can be used to optimize the efficiency and resource utilization of LLMs, help adapt the model to domain-specific vocabulary, and address bias-related issues.

5. Which capability is offered by the OCI Language service? Object Detection Speech-to-Text Conversion Image Recognition Text Sentiment Analysis

Correct Option: Text Sentiment Analysis OCI Language Service is a natural language processing (NLP) service. This service allows you to analyze text data and determine the sentiment or emotional tone expressed in the text. It identifies the sentiment of the text - and not just one sentiment for the entire block of text, but the different sentiments for different aspects.

1. What types of data are commonly used for the OCI Anomaly Detection service? Image Textual Numeric Time-Series Correct Option: Time-Series

Correct Option: Time-Series Oracle Cloud Infrastructure Anomaly Detection identifies anomalies in time-series data. Time-series data is characterized by data points collected or recorded over time, with each data point associated with a specific timestamp.

3. Which is NOT an Oracle Cloud Infrastructure AI service? Language Vision Translator Speech

Correct Option: Translator Oracle Cloud Infrastructure (OCI) offers various AI services, including Language, Speech, and Vision services. "Translator" is not a stand-alone AI service category offered by Oracle Cloud Infrastructure.

5. Which type of Machine Learning algorithms extract trends from data? Unsupervised Machine Learning Natural Language Processing Reinforcement Learning Supervised Machine Learning

Correct Option: Unsupervised Machine Learning The Unsupervised Machine Learning algorithms extract trends from data. In contrast, Supervised Machine Learning (option a) involves using labeled data to train algorithms to predict outcomes or classify data, and Reinforcement Learning (option c) focuses on training agents to make sequences of decisions through trial and error to maximize rewards. Natural Language Processing (option d) is a field within machine learning that deals with processing and understanding human language. Although NLP can be used to extract trends and insights from text data, it's not a type of Machine Learning algorithm in the same sense as the other options.

1. Which task is a Generative AI task? Calculating the total word count in an article Identifying the main topic of a news article Detecting credit card fraud Writing a poem based on a given theme

Correct Option: Writing a poem based on a given theme Writing a poem based on a given theme is an example of a Generative AI task. Generative AI refers to AI systems that can generate creative content, such as text, images, music, and more. In this case, the AI is given a theme or a prompt and then generates a new piece of creative writing, which is the poem.

4. What is the goal of prompt engineering in the context of Large Language Models? Fine-tuning model hyperparameters Optimizing hardware infrastructure for model training Crafting of specific instructions or queries for the model Automating of data preprocessing tasks

Crafting of specific instructions or queries for the model Prompt engineering involves designing specific prompts, instructions, or queries that guide Large Language Models to generate desired responses or perform specific tasks.

What is the primary purpose of deep learning model architectures like convolutional Neural Networks (CNNs)? Generating high-resolution images. Creating music compositions. Processing sequential data. Detecting patterns in images.

Detecting patterns in images. Explanation: Convolutional Neural Networks (CNNs) are specifically designed to process and analyze visual data, such as images. They excel at detecting patterns, features, and objects within images.

9. Which is NOT a part of Oracle Responsible AI guidelines? Developing policies and procedures Ensuring compliance Setting up governance Ensuring equality

Ensuring equality The responsible AI requirements involve setting up governance, developing policies and procedures, and ensuring compliance.

In Machine Learning, what does the term "Model training" involve? Writing code for the entire program. Collecting and labeling data. Establishing a relationship between input and output parameters. Analyzing the accuracy of a trained model

Establishing a relationship between input and output parameters. Explanation: Model training involves building a relationship between the input features and the desired output. It's the process of creating a model that can make predictions based on input data.

What is the primary distinction between generative AI and other AI approaches like supervised learning? Generative AI focuses on decision-making and optimization. Generative AI aims to understand underlying data distribution & create new examples. Generative AI generates labeled outputs for training. Generative AI is exclusively used for text-based applications.

Generative AI aims to understand underlying data distribution & create new examples. Explanation: Generative AI goes beyond making predictions or decisions. It focuses on modeling the structure of the data and creating new examples that resemble the training data, allowing for the generation of new content.

1. Which component of feedforward neural network is responsible for processing input data and forwarding it through various layers to make predictions? Input Layer Hidden Layer Bias Layer Output Layer

Input Layer In a feedforward neural network, the input layer is responsible for processing the input data and forwarding it through the network. It acts as the initial layer that receives the raw input features, such as pixel values in an image, and passes this information to the subsequent hidden layers for further processing and, ultimately, to the output layer for making predictions.

4. Which type of Recurrent Neural Network (RNN) architecture is used for Machine Translation? Many-to-Many One-to-Many Many-to-One One-to-One Correct Option: Many-to-Many

Machine Translation involves translating a sentence or a sequence of text from one language to another, which is essentially a sequence-to-sequence problem. In the Many-to-Many RNN architecture, the network takes a sequence of inputs and produces a sequence of outputs. In the context of machine translation, this means it can take a sequence of words or tokens in one language as input and generate a corresponding sequence of words or tokens in another language as output.

Which AI domain is associated with tasks like identifying the sentiment of a text and translating text between languages? Natural Language processing Computer Vision Speech Processing Anomaly Detection

Natural Language processing Explanation: The natural language processing domain of AI focuses on tasks related to understanding, processing, and generating natural language, such as sentiment analysis, translation, and text classification.

11. An online bank wants to streamline the loan approval process. They have historical data on past loan applicants, including information on applicants' credit score, annual income, employment status, and whether they repaid the loan or defaulted. Which machine learning algorithm can be used for this application? Supervised Machine learning for classification Reinforcement Machine Learning Supervised Machine Learning for regression analysis Unsupervised Machine Learning

Supervised Machine learning for classification In this scenario both the input features and the output categorial labels are provided by the historical data. You have labeled data (applicants' loan repayment status) and want to build a model that can classify new applicants as either "approved" or "not approved" based on features such as credit score, annual income, and employment status. Classification is the specific type of supervised learning where the goal is to assign data points to predefined categories or classes.

10. Which type of machine learning algorithm is used in predicting house prices? Supervised Classification Unsupervised Learning Supervised Regression Reinforcement Learning

Supervised Regression Regression is a supervised machine learning algorithm, which predicts a numerical output. In the case of house rental price prediction, the output, which is the rental price, is a numerical value.

What is the advantage of using OCI superclusters for AI workloads? They are ideal for tasks like text-to-speech conversion. They offer seamless integration with social media platforms. They provide a cost-effective solution for simple AI tasks. They deliver exceptional performance and scalability for complex AI tasks.

They deliver exceptional performance and scalability for complex AI tasks. Explanation: OCI AI Super clusters are specifically designed to handle demanding AI workloads that require significant computational power and scalability. They are optimized to provide high performance for complex tasks like training large machine learning models. Deep learning, and other compute-intense AI tasks.

12. What is the purpose of the self-attention mechanism in Transformer-based models? To model dependencies between different elements in a sequence To perform dimensionality reduction on input data To calculate feature embeddings for images To compute gradients for backpropagation

To model dependencies between different elements in a sequence The self-attention mechanism in Transformer models is primarily used to model dependencies and relationships between different elements (tokens) in a sequence, making it well-suited for various natural language processing tasks and sequence-to-sequence tasks.

What role do tokens play in large Language Models (LLMS) Tokens represent the numerical values of model parameters. Tokens determine the size of model's memory. Tokens are individual units into which a piece of text is divided during processing by the model. Tokens are used to define the architecture of the model's neural network.

Tokens are individual units into which a piece of text is divided during processing by the model. Explanation: In the context of LLMs, tokens are the individual units into which a piece of text is divided during processing. Tokens are usually words, sub words, or characters. LLMs process and analyze these tokens to understand and generate text.

Which is a feature offered by the Oracle Cloud Infrastructure (OCI) Speech Service? Converting text into images. Recognizing objects in images. Transcribing spoken language into written text. Analyzing sentiment in text.

Transcribing spoken language into written text. Explanation: OCI Speech service enables the conversion of spoken language into written text, making it useful for applications like transcription services and voice assistants.

13. Which tasks uses Generative AI? Video generation Classification of objects Predicting Stock prices Face recognition

Video generation Generative AI learns the underlying patterns in a dataset and uses that knowledge to create data and shares that pattern. In case of video generation, given a set of images, Generative AI would be able to generate a video.

2. Which AI domain is primarily associated with the task of detecting and recognizing faces in images or videos? Speech Language Vision Reinforcement learning

Vision Computer vision is the AI domain primarily associated with the task of detecting and recognizing faces in images or videos. Computer vision algorithms and models are specifically designed to process visual data, making them well-suited for tasks such as face detection and facial recognition.