from John C
These notes are what I added to my talk on ChatGPT for the OLLI class:
Quite often people will begin to think of an AI chat bot as their friend and confidant. This can lead you into relying on AI for your every day life decisions. It is best to think of AI as an assistant for finding answers and not for giving you personal or financial advice. Even the answers that AI provides need to be vetted for accuracy. Do no make the mistake thinking that all responses are real and true. With AI, you need to constantly question its “authority.” It would be good practice to give more than one AI chat bot with the same prompt and compare results. ChatGPT, Claude, and Perplexity are good choices to work with. Grok, Gemini, Meta, CoPilot, Kimi and DeepSeek are not (my opinion). The website you.com is good starting point for tinkering with different AI models all in one place (20+ models). Many AI chat bots have apps that run locally on your Mac, iPhone, and iPad. When accessing any chat bot, be sure to not select one that clearly states something like “based on …”. Those bots are created by independent developers whose agenda might be to scrape as much information about you as possible to either manipulate you or to sell to marketers. Only use AI chat bots that are provided exclusively by the prime developer, like openai.com (ChatGPT), anthropic.com (Claude), and perplexity.ai (perplexity).
Perplexity is the most reliable AI for research and fact-based queries, but not the top choice for creative or generative tasks where ChatGPT and Claude tend to lead.
AI often gives responses that reinforce our existing beliefs, and that’s because it learns how we think and what we like and what we don’t like.
In a nutshell:
🟢 ChatGPT — Best all-rounder. Ideal for coding, creative projects, image generation (DALL·E), and voice interaction. The most feature-rich ecosystem.
🟣 Claude — Best for nuanced, thoughtful conversation, long document analysis, and safety-conscious use cases. Tends to feel the most natural and human in tone.
🔵 Perplexity — Best for research and fact-finding. If you need accurate, cited, real-time answers to specific questions, it's the most trustworthy tool for that job.
The "best" AI really depends on your use case — many power users actually use all three for different tasks.
Prompt Engineering
from John C
What is prompt engineering?
Prompt engineering for AI involves crafting effective prompts or input queries to elicit desired responses from AI models. It’s about formulating questions or statements in a way that guides the model to generate the desired output. This can involve adjusting the wording, providing context, or using specific formatting to achieve the intended results. Effective prompt engineering is crucial for getting accurate and relevant responses from AI systems.
What would be an excellent AI prompt for users of AI?
An excellent general-purpose AI prompt for users—especially those new to generative AI—should be clear, specific, and guide the model toward useful output. Here’s a strong example:
“Act as a knowledgeable assistant. I want to [achieve a goal]. Provide clear, step-by-step guidance, include examples if helpful, and ask follow-up questions to refine your answer if needed.”
This structure works well because it:
Defines the AI’s role
States a goal
Asks for clear and practical help
Leaves room for interaction
Don’t be afraid to use simple prompts, like “What is a good product for repairing a hole in the imitation leather of an office chair?” Or, “How can I extend the life of plastic patio furniture?”