Generative AI in a Weekend: Learn Google Gemini

Understand the goals of artificial intelligence and the difference between strong general artificial intelligence and narrow weak artificial intelligence, with examples like self-driving cars, fraud detection, and email classification.

Explore Bard, Google's generative AI chatbot, via bard.google.com, and practice prompts like design briefs or leadership keywords, while noting Bard's limitations and variable accuracy.

Explore how Bard supports coding, learning, and design—from Python leap-year code to REST API design and relational databases, with DDL and test data, and guidance on Head First Design Patterns.

Explore Bard's use cases for technology and outline a cloud engineering learning path across cloud platforms, languages, storage, networking, security, DevOps, containers, serverless, Docker steps, and React and Angular comparison.

Explore how Bard generates prompts to understand DevOps at a high level. Dive into prompts about Docker, Jenkins, monitoring tools, and securing pipelines, tailored for beginners.

Generative AI is not deterministic. Sending the same prompt often yields a new response, so expect varying outputs, and learn how a few parameters can influence this.

Explore how Bard fits AI, ML, and generative AI, and distinguish between them, while noting limitations, use cases, and the role of large language and foundation models.

Contrast traditional programming with machine learning, showing how ML learns from millions of examples to build predictive models and distinguish strong, general, and narrow AI.

Learn how machine learning uses features as inputs and labels as outputs to make predictions, with examples like house and car prices, spam detection, and classification versus regression.

Learn the full cycle of machine learning, from obtaining and cleaning data to feature engineering, training a model, evaluating it, and deploying for production predictions.

Confront the AI turmoil pragmatically by learning to use AI APIs, integrating them into your applications, and boosting productivity without fearing the future.

Explore how generative AI sits inside artificial intelligence and machine learning, learning from examples to create new text, code, images, and videos rather than only predict outcomes.

Explore how generative ai requires huge volumes of data, training models on diverse datasets, including images, text, code, books, and forums, to generate new data.

Generative AI uses self-supervised learning, training on large volumes of text without explicit labels by predicting the next word and learning from mistakes. It builds contextual understanding from predicted words.

Explore how generative AI for text predicts the next word from context and learn how temperature, top_k, and top_p control the chosen word.

Explore how generative AI text models use tokens instead of words to maintain consistency, learn relationships between tokens and parts of speech, and manage token limits by splitting long outputs.

Compare predictive machine learning with generative AI by examining inputs, outputs, use cases, and prompts that generate new content.

Explore Google Cloud's traditional ML landscape before the emergence of generative AI, from API-based options with no training to AutoML in Vertex AI and custom training.

Explore Google Cloud's machine learning API suite, including natural language, speech-to-text, text-to-speech, vision, and video intelligence APIs. Derive insights from unstructured text, detect entities and sentiment, and enable content moderation.

Build custom machine learning models without expertise using AutoML in Vertex AI. Label images, text, tables, and videos to train models that predict cloud types, sentiment, and house prices.

Use Vertex AI custom training to run your own code in a container, tune hyperparameters, and build complex models with TensorFlow, PyTorch, or scikit-learn using BigQuery ML on BigQuery data.

Create a Google Cloud account at cloud.google.com, claim the $300 free credit, verify your mobile number, and enter card details to start your free trial.

This quick Google Cloud Vertex AI demo guides you to create a project, build an image classification dataset, label images, and train a model with AutoML or custom training.

Discover Google's generative AI landscape with Vertex AI, Model Garden, and foundation models across language, vision, tabular, document-based, and speech, plus tools like Generative AI Studio and MakerSuite.

Explore foundation models in Vertex AI Model Garden, including text-bison, chat-bison, and textembedding-gecko. Learn about code-bison, codechat-bison, code-gecko, imagegeneration, and imagetext for code, chat, and image tasks.

Explore text models such as text-bison, chat-bison, and textembedding-gecko to perform summarization, classification, and content creation, including generating hashtags and course titles from texts.

Explore text classification and sentiment analysis in Generative AI, turning feedback, news, and reviews into labeled categories such as positive, negative, or neutral. See examples like politics, sports, entertainment.

Explore text features with extraction and writing, including JSON extraction and testing prompts. Learn how structured and freeform layouts organize examples and prompts for effective evaluation.

Explore ideation with text features to generate ideas, names, advice, interview questions, and trends for cloud and DevOps, including cost management best practices.

Reframe failures as learning opportunities, analyze what went wrong, keep a journal, and seek mentor or peer input for diverse perspectives while embracing hands-on learning.

Design effective prompts to elicit high-quality, reliable language model responses. Learn best practices: clear instructions, examples, experimentation, and frameworks such as RTF, CTF, and RASCEF.

Explore prompt design by comparing zero-shot, one-shot, and many-shot approaches, using concrete examples and predefined answer formats to improve accuracy and consistency when returning structured outputs like JSON.

Explore prompt frameworks—RTF, CTF, and RASCEF—and learn how to craft clear prompts with defined role, task, format, or context, to produce precise responses.

Experiment with parameters in the Generative AI Studio to control output length, model choice, and randomness, using max token limit and temperature, top-K, and top-P.

Learn to execute the Vertex AI PaLM API from a Vertex AI workbench notebook in Google Cloud, using the google-cloud-aiplatform library, project ID, location, and vertexai.init.

Learn to call Vertex AI language models from your app using the PaLM text model API, with and without examples, by writing Python code, executing prompts, and running notebooks.

Explore generative AI chat models using chat-bison on Vertex AI’s Model Garden PaLM for chat, with text chat and context-driven conversations for customer service and education.

Explore foundation code models such as code-bison, codechat-bison, and code-gecko to generate, explain, test, and containerize code with human-involved, safe practices, using Vertex AI for prompts.

This step introduces image foundation models like Imagen for text-to-image generation, image captioning, and visual Q&A within Vision, including uploading images, generating captions, editing images, and answering image questions.

Explore the Chirp-based speech model and master text-to-speech and speech-to-text capabilities, with practical code to automate these features via APIs.

Manage costs in cloud notebooks by stopping unused user-managed notebooks. Restart when needed and delete the notebook at the end of the course to minimize charges.

Tune a language model to improve accuracy when prompt engineering falls short, using a JSONL training dataset of hundreds of input_text and output_text examples in Vertex AI Generative AI Studio.

Explore embeddings as vector representations in a high-dimensional space to capture semantic relationships and contextual information, enabling text similarity, recommendation systems, clustering, and outlier detection in natural language processing.

Explore embeddings by performing a similarity search to find the most similar sentence to a given input using a dot-product measure in a pandas dataframe, with Vertex AI text embeddings.

Learn how LangChain enables easy access to multiple language models, using text, chat, and embedding abstractions, and build chains to combine tools and data sources.

Master LangChain fundamentals by building prompts for Vertex AI PaLM using LLM interfaces, chat models, and embeddings. Use PromptTemplate and simple sequential chains to compose multi-step tasks.

Learn to answer questions from long articles with LangChain by splitting text with RecursiveCharacterTextSplitter into chunks, using embeddings and FAISS for similarity search, then querying a PaLM model.

Explore LangChain summarization using single-prompt and map-reduce approaches, splitting multiple documents, summarizing individually with Vertex AI PaLM API, and producing a final combined summary.

Learn how to run Vertex AI PaLM API from Colab by installing google-cloud-aiplatform and restarting the kernel. Authenticate to Google Cloud and initialize Vertex AI with project ID and location.

Explore Generative AI App Builder and understand why access may show documentation only and how general availability is limited to allowlisted customers. Join the Trusted Tester program for early access.

Learn how Gen App Builder enables building enterprise-grade generative AI apps without code, including search, chat, and recommendations, using data from websites, unstructured documents, or structured data in data stores.

Explain makerSuite and PaLM API availability, noting they are not GA yet and access is via a waitlist. Register for the waitlist for a chance at early access.

Learn to use PaLM API and MakerSuite to prototype Google's generative AI without Google Cloud. Access text, chat, and data prompts with an API key and run prompts in Colab.

Back up your work from the user-managed notebook in Vertex AI Workbench by downloading files in JupyterLab before deleting the notebook, to avoid losing access to your labs.