<ul><li class="ql-align-justify">Takes natural language input and produces text output</li><li class="ql-align-justify">Learns to map one text to another</li><li class="ql-align-justify">Common use cases:</li></ul><ol><li class="ql-align-justify">Translation (e.g., English ↔ Spanish)</li><li class="ql-align-justify">Summarization</li><li class="ql-align-justify">Text generation (e.g., writing assistance)</li></ol>

📝 Text-to-Text

Helps build AI-powered search and conversational agents for both customers and internal teams.Use Cases:<ul><li class="ql-align-justify">🤖 Chatbots</li><li class="ql-align-justify">🧠 Digital assistants</li><li class="ql-align-justify">🔍 Custom search engines</li><li class="ql-align-justify">📚 Knowledge bases</li><li class="ql-align-justify">🎓 Training and learning applications</li></ul>Ideal for enhancing customer experience and internal workflows through intelligent, conversational interfaces.

💬 Vertex AI Agent Builder (formerly Vertex AI Search and Conversation)

<ul><li class="ql-align-justify">Supervised Learning (Labeled Data)</li><li class="ql-align-justify">Data provided includes tags</li><li class="ql-align-justify">Learns from past examples to predict future outcomes</li><li class="ql-align-justify">Example: Predict amount of tips gave customer based on order type</li></ul>

Machine Learning Model - Supervised

<ul><li class="ql-align-justify">Generates 3D models from a user’s text description</li><li class="ql-align-justify">Used in game design, virtual worlds, or AR/VR applications</li><li class="ql-align-justify">Example: "A futuristic flying car with neon lights"</li></ul>

📦 Text-to-3D

<ul><li class="ql-align-justify">A subfield of AI</li><li class="ql-align-justify">Uses input data to train models</li><li class="ql-align-justify">Trained models predict unseen but similar data</li><li class="ql-align-justify">Enables computers to learn without explicit programming</li></ul>

What is Machine Learning?

<ul><li class="ql-align-justify">This approach uses training code and labeled data to build a model. Depending on the task, the model can:</li><li class="ql-align-justify">Make predictions</li><li class="ql-align-justify">Classify inputs</li><li class="ql-align-justify">Cluster similar data points</li></ul>🐱 Before ML: Rule-Based Programming<ul><li class="ql-align-justify">In traditional programming, we had to hardcode rules to define what a cat is:</li><li class="ql-align-justify">Type: animal</li><li class="ql-align-justify">Legs: 4</li><li class="ql-align-justify">Ears: 2</li><li class="ql-align-justify">Fur: yes</li><li class="ql-align-justify">Likes: yarn, catnip</li><li class="ql-align-justify">Dislikes: Fred</li></ul>

🧠 Traditional Machine Learning (Supervised Learning)

<ul><li class="ql-align-justify">Uses artificial neural networks to learn complex patterns</li><li class="ql-align-justify">Neural networks are inspired by the human brain (made of many interconnected neurons)</li><li class="ql-align-justify">Deep learning models have many layers of neurons</li><li class="ql-align-justify">Can work with supervised, unsupervised, or semi-supervised data</li><li class="ql-align-justify">Semi-supervised = small labeled data + large unlabeled data</li><li class="ql-align-justify">Labeled data teaches core concepts; unlabeled data helps generalize</li><li class="ql-align-justify">Generative AI and Large Language Models (LLMs) are subsets of deep learning</li></ul>

Deep Learning (Subset of Machine Learning)

<ul><li class="ql-align-justify">Generates videos from text input</li><li class="ql-align-justify">Input can range from a simple sentence to a detailed script</li><li class="ql-align-justify">Example: "Show a sunrise over the mountains with birds flying by"</li></ul>

🎥 Text-to-Video

A platform that enables developers to quickly explore, customize, and deploy generative AI models on Google Cloud.Key Features:<ul><li class="ql-align-justify">📚 Library of pre-trained models</li><li class="ql-align-justify">🛠️ Tools for fine-tuning</li><li class="ql-align-justify">🚀 Deployment to production</li><li class="ql-align-justify">💬 Community forums for collaboration and idea sharing</li></ul>Great for developers looking to get started with generative AI using ready-to-use resources.

🧰 Vertex AI Studio

<ul><li class="ql-align-justify">Generative AI models take in training code, labeled data, and even unlabeled data of all kinds (text, images, audio, etc.) to build a powerful foundation model.</li><li class="ql-align-justify">This model can then create brand new content — not just recognize patterns but generate text, code, images, audio, and video.</li></ul>📸 Neural Networks in Action<ul><li class="ql-align-justify">With neural networks, we can now show the system images of cats and dogs and ask:</li><li class="ql-align-justify">“Is this a cat?”</li><li class="ql-align-justify">The model doesn't need hardcoded rules — it learns from data and makes a prediction: cat or not a cat.</li></ul>

🤖 Generative AI: A New Era

Predictive ML Model VS Gen AI Model (Chart)

Predictive ML Model VS Gen AI Model

Goal: Create new content (text, images, etc.)How: Learns joint probability p(x, y) and generates new data based on patternsExample: Generate a dog image based on understanding what a "dog" looks like

Deep Learning Model (Generative)

<ul><li class="ql-align-justify"> A discipline of computer science</li><li class="ql-align-justify">Creation of autonomous, intelligent agents that reason and learn.</li><li class="ql-align-justify">Theory and methods to build human-like thinking and acting machines.</li></ul>

What is Artificial Intelligence?

<ul><li class="ql-align-justify">Driven by transformers, a deep learning architecture</li><li class="ql-align-justify">Key breakthrough in 2018 revolutionized Natural Language Processing (NLP)</li><li class="ql-align-justify">Transformers enable GenAI to understand and generate contextually relevant content</li></ul>

⚙️ Power Behind GenAI

How to differentiate Gen AI?

<ul><li class="ql-align-justify">Takes images as input</li><li class="ql-align-justify">Can output images, text, or videos</li><li class="ql-align-justify">Examples: image captioning, super-resolution, video generation, and image completion</li></ul>

🖼️ Generative Image Model

<ul><li class="ql-align-justify">Trained on large datasets of images with text descriptions</li><li class="ql-align-justify">Generates realistic or stylized images from a text prompt</li><li class="ql-align-justify">Diffusion models are often used in this process</li><li class="ql-align-justify">Example: "A cat wearing sunglasses riding a skateboard"</li></ul>

🖼️ Text-to-Image

<ul><li class="ql-align-justify">Goal: Classify or label data</li><li class="ql-align-justify">How: Learns conditional probability p(y | x) to predict labels for new inputs</li><li class="ql-align-justify">Example: Given an image, predict if it's a dog or a cat </li></ul>

Deep Learning Model (Discriminative)

Generative AI can sometimes produce hallucinations — content that sounds plausible but is incorrect or nonsensical.Reasons include:<ul><li class="ql-align-justify">Insufficient training data</li><li class="ql-align-justify">Poor-quality (noisy or dirty) training data</li><li class="ql-align-justify">Lack of context in the prompt</li><li class="ql-align-justify">Inadequate constraints or guidance in prompt design</li></ul>

⚠️ Common Issues with Generative AI

🚀 Generative AI Applications

🌱 Generative AI Model Types Explained

🧠 How Generative AI Works

A prompt is a short input text given to a Large Language Model (LLM) to guide its output.<ul><li class="ql-align-justify">Example: “Summarize this article,” or “Write a poem in the style of Shakespeare.”</li><li class="ql-align-justify">Prompt design is the art of crafting effective prompts to achieve desired outcomes from the model.</li></ul>

✏️ Prompting

Generative models create new data, while discriminative models learn to tell data apart.

<ul><li class="ql-align-justify">No tags or labels on the data</li><li class="ql-align-justify">The goal is to explore and find patterns or groups in the raw data</li></ul>

Machine Learning Model - Unsupervised

<ul><li class="ql-align-justify">Takes text as input</li><li class="ql-align-justify">Can generate more text, images, audio, or even make decisions</li><li class="ql-align-justify">It learns language patterns from training data</li><li class="ql-align-justify">Predicts how to complete or respond to a sentence</li></ul>

📝 Generative Language Model

Gemini goes beyond text and enables multimodal capabilities, including:<ul><li class="ql-align-justify">🖼️ Image analysis</li><li class="ql-align-justify">🔊 Audio understanding</li><li class="ql-align-justify">💻 Code interpretation</li></ul>A versatile tool for tasks requiring visual, audio, and programming intelligence — all from a single model.

🌐 Gemini

<ul><li class="ql-align-justify">Executes specific tasks based on text input</li><li class="ql-align-justify">Can perform actions like:</li></ul><ol><li class="ql-align-justify">Answering questions</li><li class="ql-align-justify">Running searches</li><li class="ql-align-justify">Making predictions</li><li class="ql-align-justify">Automating workflows (e.g., navigating UIs, editing documents)</li></ol>

⚙️ Text-to-Task

Generative AI (GenAI) is a type of Artificial Intelligence that creates new content based on patterns it has learned from existing data.<ul><li class="ql-align-justify">The learning process is called training, which produces a statistical model.</li><li class="ql-align-justify">When given a prompt, GenAI uses this model to predict likely responses — generating new content.</li><li class="ql-align-justify">It captures the underlying structure of the data and can produce similar, but original, outputs.</li></ul>

🔍 What is Generative AI?

<ul><li class="ql-align-justify">Large-scale models trained on massive datasets</li><li class="ql-align-justify">Adaptable to many downstream tasks (via fine-tuning or prompt engineering)</li><li class="ql-align-justify">Tasks include:</li></ul><ol><li class="ql-align-justify">Sentiment analysis</li><li class="ql-align-justify">Image captioning</li><li class="ql-align-justify">Object recognition</li><li class="ql-align-justify">Question answering</li></ol><ul><li class="ql-align-justify">Industry Impact:</li></ul><ol><li class="ql-align-justify">Healthcare, finance, customer service, fraud detection, and more</li></ol><ul><li class="ql-align-justify">Example platform: Vertex AI’s Model Garden with support for both:</li></ul><ol><li class="ql-align-justify">Language models (chat, code, summarization)</li><li class="ql-align-justify">Vision models (image captioning, embeddings, diffusion models)</li></ol>

Predictive ML Model VS Gen AI Model

Similar ideas to Predictive ML Model VS Gen AI Model

Sculpting vs. Gardening: Tools for Collaboration

The ML workflow

Explanations vs. Predictions vs. Generations