AI Essentials : Concepts and Demos

Identify the problem, assess traditional methods, define the desired AI output, collect and preprocess data, select and train models, and pursue deployment with continuous learning.

Assess whether the problem can be solved by traditional IT or AI, and define hypotheses or assumptions, scope and limitations, data, and success criteria, illustrated by an insurance claims-driven chatbot.

Identify data collection methods and tools for internal and external data, ensuring data quality, integrity, and integration across formats like text, numbers, images, video, and audio to train ai models.

Process raw data through data smoothing, clean missing values and errors, transform into formats for multi-modal inputs, and extract relevant fields to train AI models.

Choose a model by weighing problem complexity, data type, and computational cost, plus ethical considerations and integration with existing systems. Learn the three model types: supervised, unsupervised, and reinforcement learning.

Identify the problem and data, split into training and validation sets, train and validate the model, then evaluate accuracy, precision, and recall while checking data quality and bias.

Optimize models to boost operational efficiency and effectiveness by reducing CPU and memory usage while improving accuracy. Apply source-code changes and regularization to address underfitting and overfitting by adjusting weights.

Deploy your trained model into a real-world system, monitor results, and feed new data to improve accuracy. Continuously learn by fine-tuning with data and leveraging human feedback and reinforcement learning.

Explore applying ai to real world problems by following its workflow: define the problem, assess ai viability, collect and clean data, select and deploy models, and evaluate results.

Explore how an insurance company can implement an automated chat bot that understands voice or text, identifies customer intent, analyzes data, and delivers human-like, cost-effective responses.

Explore how Amazon Lex enables building, testing, and deploying chatbots on websites including Facebook Messenger, with machine learning for initial bot design, templates, and seamless, cost-effective deployment.

Explore building a chatbot using AWS S3 and Amazon Lex, creating S3 buckets and subfolders, uploading sample data, analyzing transcripts to generate intents, and deploying the bot.

Build an insurance chatbot as a practice exercise using Amazon Lex and Amazon Connect, with Lambda for business logic and Kendra for queries, all secured by IAM roles.

Explore generative ai and discriminative ai, comparing content creation with data classification, and examine how unsupervised and supervised learning shape training data and model outputs.

Foundational models provide a starting point for rapid, cost-effective ML by training on broad data and adapting to diverse tasks like NLP, language translation, image generation, and content creation.

Explore ChatGPT, an AI chatbot from OpenAI built on the GPT-3 family, now featuring GPT-4 access. Learn how to access it via chat.openai.com, the old chat.openai.com/chat, or the official app.

Explore a practical ChatGPT demo with prompts, pricing plans, and model options. Learn to tailor prompts, generate Python code, write letters, and craft emails using free GPT-3.5 and plus GPT-4.

Explore Amazon Q for business conversations and insight generation, from support tickets to policy information, alongside CodeWhisperer as an AI code companion for generating starter code.

Watch an end-to-end demo of Amazon Q and Codewhisperer, log in to the AWS console, install the AWS toolkit, and explore code generation and conversion capabilities.

Explore google bard, google's experimental conversational ai that uses palm two, pulls data from google search, supports many languages, and uses multimodal input including images.

Explore a google bard demo, including sign-in, prompts, and live search results. See how it handles coding prompts, image outputs, and arithmetic limitations.

Explore prompt engineering fundamentals, define prompts and their elements. Learn techniques to guide AI models to produce desired results, considering model choice and data.

Discover how a prompt guides AI to produce desired outputs by providing precise instructions and keywords, illustrated through a plumber’s seven-year journey from poverty to influence.

Explore how prompt design shapes AI output by detailing instruction, context, input data, and output indicators, with five effective principles: clarity, context, precision, persona, and pattern.

Master prompt engineering by defining prompts and their elements, and applying techniques to guide generative AI to produce desired outputs, considering model selection and training or validation data.

Design the intended output and tasks in the prompt life cycle, then implement, evaluate, refine, iterate, validate, deploy, and maintain robust AI outputs.

Master prompt engineering to optimize efficiency, boost task performance, and understand model constraints for accurate outputs, using tools like IBM Watson AI Prompt Lab, Spell book, dust, and prompt perfect.

Explore text-to-text and text-to-image prompt techniques for reliable, unbiased prompts in LMS, covering objective setting, contextual guidance, domain terminology, and bias mitigation.

Explore zero-shot and few-shot prompting, learn how to provide instructions with or without examples, and apply chain-of-thought reasoning to solve multi-step problems.

Learn the least-to-most prompting technique, building on tree of thought by solving subproblems first and using outputs to tackle the bigger problem. Use a 15-minute time example to count trips.

Explore text-to-text prompting techniques, including complexity-based prompting with chain-of-thought rollouts, generated knowledge prompting, self-refined prompting, and directional stimulus prompting. Use prompts with hints to steer model outputs toward desired result.

Explore text-to-image prompts and how AI models generate images from instructions, shaping quality, relevance, and diversity with style modifiers, size, color, contrast, texture, and artistic references.

Boost image quality by using prompts with blur background, sharp details, and 4K or 8K resolution; apply repetition and weighted terms to emphasize elements like a dense forest and sparrow.

Explore the foundations of ethics in AI, define AI principles, and examine ground rules, ethics pillars, and regulations to ensure AI decisions align with human morals.

Define ethics in AI, emphasizing responsible, transparent, and explainable practices across the AI lifecycle to maximize benefits, protect data ownership, and curb risks like deepfake misuse.

Explore the ethics pillars, including fairness, reliability, privacy, transparency, sustainability, and explainability, and how they guide AI policy, prevent bias by gender, color, or race, and protect human well-being.

Examine the challenges and benefits of AI regulation, from pace of development and scope to ethical use, transparency, fairness, and references to the AI Act and NIST risk management framework.