Complete ChatGPT Prompt Engineering Course

This lecture explains artificial narrow, general, and superintelligence—ANI, AGI, and ASI—and how humanity sits on the spectrum toward singularity. It shows how to interact with AI by writing good prompts.

Learn how to speak the same language as AI by comparing programming languages, understanding prompt quality (garbage in, garbage out), and applying clear, structured prompts with context and evaluation strategies.

Use a crafted ChatGPT prompt to reply to Lucy's email, paste the message, set context, and amend with a skeleton to shorten by 25% and remove after sales service lines.

Explore natural language processing and AI language models, including left-to-right autoregressive, masked, encoder-decoder, and multimodal models, with GPT-3.5 and GPT-4 in real exam demonstrations.

Explore how gpt models use transformers to learn from large data, through tokenization, embeddings, positional encoding, multi-head self-attention, and next-word predictions.

Discover how ChatGPT uses long-term memory through pre-training and fine-tuning, and how short-term memory, context length, and token limits shape responses and chat history.

Explore zero-shot, one-shot, and few-shot prompting in AI, from relying on pre-trained knowledge to using examples to shape outputs. Learn use cases when data is scarce or quick responses matter.

Learn query decomposition, or incremental query refinement, by breaking prompts into key components using a tree diagram, enabling more specific, compilable responses for tasks like writing an ebook.

Learn decision making in prompt design, comparing intuitive prompting with prompt engineering. Explore structure, variables, context, and evaluating responses, and decide when quick answers suffice or full prompts are needed.

Explore the pitfalls of LLMs, including hallucinations, bias, token limits, math limits, and data cutoff. Apply validation, response analysis, and tools like Wolfram Alpha, embeddings, and mitigations for prompt hacking.

Discover multimodal updates to ChatGPT plus, enabling image analysis and DALL-E 3 image generation. Learn to upload PNG or JPEG, craft prompts, and refine logos while avoiding text in images.

Follow along with the projects to build prompts using ChatGPT and Playground, applying the theory’s tools and techniques and preparing for assignments after each chapter.

Create a high quality LinkedIn about summary from your credentials using a Reid Hoffman style prompt, and learn hyperparameters, model selection, zero-shot prompting, temperature and top_p, and length control.

Explore hyperparameters and their role in shaping model behavior within OpenAI's Playground, and learn the science and art of prompt engineering to iteratively refine prompts for better ai responses.

Choose among four modes in OpenAI's playground, focusing on complete mode and the completions API, with presets and a chat version using system, user, and assistant messages for continuous prompting.

Learn to use token probabilities to debug and influence AI text generation, toggling most likely, least likely, and full spectrum options, and adjust temperature and top P.

Learn to control computational creativity by adjusting temperature and top P, balancing accuracy and creativity while iterating to fine-tune output quality.

Master the maximum length setting by linking token limits (2048 output, 4000 total) to character counts, applying a single character cap for LinkedIn summaries.

Learn how frequency penalty reduces repeated phrases by penalizing tokens based on prior frequency, and how presence penalty controls the creation of new concepts to shape the model's output.

Learn how to use OpenAI's chat mode in the playground, including system, user, and assistant messages, model selection, and hyperparameters to craft a blog about AI's impact.

Craft a precise system message for role-based prompting, shaping the AI as an esteemed blog writer with strong writing, research, and market skills, and a passionate audience focus.

Explore few-shot prompting and prompt engineering by refining system messages, assigning roles, and adding training examples to tailor AI blog outputs for a specific audience.

Learn to optimize your blog for search engines by applying humor, emotive language, and personal anecdotes to improve content creation, indexing, ranking, and marketing.

Assess ai content detectors and learn practical rewriting steps using content at scale, copy leaks, and QuillBot paraphrasing to produce more human-sounding, credible text.

Explore advanced system message usage to control a chat bot, embedding roles and prompts, and perform stepwise tasks like summarization and translation in a web-based OpenAI API app.

Explore custom instructions in ChatGPT Plus to tailor responses, enable new-chat presets, and apply role prompting and answer engineering with examples like lesson plans, code, and grocery lists.

Learn to build a PowerPoint presentation generator prompt that writes VBA code and auto-generates slide content and visuals, using natural language prompts and text-to-image AI.

Learn how the system message carries greater weight than user prompts, using CEO and supervisor analogies to control AI behavior and apply this to tasks like building a PowerPoint.

Iterate a PowerPoint deck with GPT by confirming output, changing text to white, inserting a GPT models slide, and using designer AI to style it, remember to fact-check.

Develop a currency converter prompt that writes Python code and uses Bing to fetch exchange rates, enabling conversion between currency pairs while mastering continuous prompting, response analysis, and code validation.

Explore response analysis by prompting, receiving, and evaluating AI outputs, learn to fact-check and proofread to avoid hallucinations, and apply optimization techniques for reliable, efficient writing.

Explore how AI language models hallucinate, caused by limited training data, prompt ambiguity, and the accuracy-fluency tradeoff, and learn practical mitigation through response analysis, refinement, and feedback.

Compare information from PWC tax summaries with SARS sources, validate turnover tax details in South Africa, and use Google Scholar to confirm accuracy, teaching practical prompt engineering against hallucinations.

Explore how user feedback trains ChatGPT through reinforcement learning with human feedback, using thumbs up or down to guide model improvements and curb hallucinations.

Validate web-scraped currency data by verifying sources against xe.com, and test all conversions from zar to usd using xe.com links before finalizing code.

Leverage your domain knowledge to set the standard for artificial intelligence prompts through education, training, and experience, and quickly spot inaccuracies in chemical engineering and heap leaching.

Discover how GPTs extend ChatGPT into task-specific agents and how to create and configure your own GPT using the three pillars—instructions, knowledge, and actions, including uploading files and DALL-E icons.

Build a custom GPT, a large action model, that answers FAQs from an Excel spreadsheet or Google Sheets, updates the sheet, and emails responses via a human-in-the-loop workflow.

Master how to create a Google Sheets row using Zapier API actions, map a question and answer, configure the spreadsheet and worksheet, and preview before testing.

Learn to configure a Gmail reply to email action, choose the account, set the thread and body with the answer, enable the action, copy the configuration, and share to publish.

Build an Amazon sales chatbot, Amazon GPT, that identifies products and provides purchase links via OpenAI API. Learn pre programming chats, prompt hacking controls, cost-efficient prompting, and prompt consolidation.

Apply shot prompting to provide context and enable in-context learning in chatbots. Craft prompts to uncover user needs, like video editing use, and steer relevant product suggestions.

Cementing characterization borders by simulating a new user with triple-bracket delimiters for context, testing prompts, and enforcing role boundaries via system messages, rules, and few-shot prompting in a sales chatbot.

Design and deploy a human-like ecommerce chatbot for amazon.com by building a structured system prompt with role and expertise, and using shot prompting and testing to funnel shoppers toward purchases.

Train a chatbot using shot prompting to keep it in character and within scope. Adjust top P to 0.8, test responses, and streamline prompts and system message to reduce cost.

Tune prompts to extend conversations by enforcing system rules that require asking up to 10 questions before suggesting a product and ending each reply with a question.

Finalize the chatbot by refining shot prompts and system messages to reduce prompts and improve accuracy for product recommendations, while safely handling sensitive topics.

Create a classic Pong game in Python using PyCharm and PyGame, leveraging one-click prompts and WebChatGPT for coding, then add scoring and easy to hard difficulty levels.

Create an auto reply email generator using a playground chat setup, leveraging shot prompting with an email thread to generate consistent assistant responses and system messages.

Explore how to convert Python to Java using ChatGPT, draft prompts with variables and delimiters, and troubleshoot common errors in Replit to generate email generator code.

Prompt ChatGPT to generate VBA code that creates a line graph from monthly profit data in Excel, using explicit cell ranges, then paste and run it from the developer ribbon.

Learn to use gpt for sheets in Google Sheets to identify the popular extracurricular activity and students from New York who major in math, using prompts, values, and api keys.

Learn to scrape the World Economic Forum Future of Jobs report 2023 PDF using ChatGPT tools and extract key skills for job security, including analytical thinking and AI skills.

Demonstrate two use cases, summarization and translation, using link reader to summarize a Time article on AI jobs and translate it into Chinese.

Learn how to download and run GPT-2 locally using Hugging Face, by installing Python, PyTorch, Transformers, and setting up a virtual environment with PyCharm to run the model offline.

Learn to fine-tune a large language model by preparing a csv dataset, converting it into prompt completion pairs, and training a custom model.

Enable the code interpreter in ChatGPT plus to write and run Python code. Discover debugging by uploading a PyCharm project, fixing errors, and downloading a corrected script.

Explore how OpenAI embeddings convert text into numerical vectors for efficient search, clustering, and similarity analysis, with practical demos in Python, tokenization, and vector visualization.