The Transformer Architecture
The Transformer Architecture: This is the "brain" behind GenAI. You must know the Attention Mechanism, which allows the model to focus on the most relevant words in a sentence regardless of how far apart they are.
2. The OCI Generative AI Service
Oracle provides a fully managed service that lets you integrate LLMs into your apps. For the exam, remember these three providers/models OCI supports:
Meta: Llama models (great for chat and general tasks).
Cohere: Command, Embed, and Rerank models (excellent for business tasks and RAG).
Mistral: High-performance open-source models.
3. Key OCI GenAI Features (High Probability for Exam)
Oracle differentiates itself through security and performance. Memorize these:
Dedicated AI Clusters: Instead of sharing hardware with other companies, OCI gives you dedicated GPU resources. This ensures consistent performance and ensures your data never leaves your environment to train public models.
Fine-Tuning: The process of taking a pre-trained model and giving it a small amount of your own data to make it an expert in a specific area (like medical or legal terms).
RAG (Retrieval-Augmented Generation): This is the most popular way to use GenAI in 2025. It connects the LLM to a private data source (like a PDF or database) so the AI can give accurate, up-to-date answers without hallucinations.
4. Oracle 23ai Integration
The 2025 exam focuses heavily on how GenAI works with the Oracle Database:
AI Vector Search: This allows the database to store "vectors" (mathematical representations of meaning). This is what powers RAG.
Select AI: A feature that allows a user to ask a question in plain English (e.g., "Show me total sales for last month") and the AI automatically converts it into a SQL query to get the data.
Comparison Table for the ExamTermWhat it doesWhy it mattersPrompt EngineeringCrafting the input to get better output.Cheapest way to improve results.In-Context LearningGiving the AI examples within the prompt.No training required.Fine-TuningUpdating the model's weights with new data.Expensive but very specialized.RAGGrounding the AI in "real-time" external data.Reduces hallucinations.
Quick Check Question:
Scenario: A bank wants to use a Large Language Model to summarize financial reports, but they are worried that their sensitive data might be used to train the public model. Which OCI feature should they use?
A) Public API
B) Dedicated AI Clusters
C) Prompt Engineering
D) OCI Vision
(The answer is B. Dedicated AI Clusters provide the privacy and isolation enterprises need.)
1. Artificial Intelligence (AI): The Umbrella2AI is the broadest category.3 It refers to any technique that enables computers to mimic human behavior or intelligence.4+1Key Concept: It doesn't always have to "learn" from data.5 Early AI used Rule-Based Systems (If-Then statements).6+1Goal: To simulate human cognitive functions (reasoning, problem-solving, perception).7OCI Exam Tip: Know that AI can be "Narrow" (specific tasks, like what we have today) or "General" (human-level, still theoretical).82. Machine Learning (ML): The Subset of AIML is a specific way of achieving AI.9 Instead of writing manual rules, we give the computer data and let it find its own patterns.10+1Key Concept: Learning from experience (data) without being explicitly programmed for every scenario.11Data Types: ML typically works best with Structured Data (tables, spreadsheets).The Three Pillars for your Exam:Supervised Learning: Learning from labeled data (e.g., predicting house prices using Regression).12Unsupervised Learning: Finding hidden patterns in unlabeled data (e.g., Customer Clustering).13Reinforcement Learning: Learning through trial and error via "rewards" (e.g., an AI agent playing a game).143. Deep Learning (DL): The Subset of MLDeep Learning is a specialized type of ML inspired by the human brain.15 It uses Artificial Neural Networks (ANN) with many "hidden" layers.16+1Key Concept: It automates "Feature Extraction."17 In traditional ML, a human might tell the AI to look at "car wheels" to identify a car. In DL, the model learns to identify those features on its own.18+1Data Types: DL excels at Unstructured Data (images, video, audio, text).19Key Models to Know:CNN (Convolutional Neural Networks):20 Used for images/Vision.RNN/LSTM (Recurrent Neural Networks):21 Used for sequences like text or speech.22+1Transformers: The architecture behind Generative AI and LLMs.Comparison for the ExamFeatureArtificial IntelligenceMachine LearningDeep LearningRelationshipThe whole umbrella.A subset of AI.A subset of ML.LogicRules or Algorithms.Statistical patterns.Neural Networks (Layers).Data RequirementLow to High.Moderate (Structured).Huge (Unstructured).Feature EngineeringManual.Manual/Human-led.Automatic (the model does it).23OCI Exam Scenario PracticeQuestion: A developer is building a system to automatically extract text from scanned medical invoices (unstructured data) using OCI Document Understanding. Which category of technology is primarily being used?A) Rule-Based AI24B) Unsupervised Machine Learning25C) Deep Learning26D) Linear Regression27Answer: C (Deep Learning). Tasks involving images, vision, and complex text extraction are almost always powered by Deep Learning models (like CNNs or Transformers).28Difference Between AI, Machine Learning and Deep LearningThis video provides a clear, visual explanation of the "Matryoshka doll" relationship between these three fields, which is essential for correctly identifying them on your OCI certification exam.
Comments
Post a Comment