Get the answers you want from AI
(Post to be updated weekly to help you build your expertise).
Quick background - How does AI work?
What are LLMs? Large Language Models are specialized programs that have been fed as much available knowledge on the internet as possible and they’ve chunked that knowledge into a very specific kind of database that’s best at storing domains of knowledge and the patterns of words and phrases that are used to describe something in a domain.
What’s made them all of a sudden “appear” on the scene in the last couple of years, and feel human-like in their responses is that researchers and scientists figured out a way to mathematically move efficiently through these large databases in and between knowledge domains and also how to apply probabilities to the most likely words in those domains so they became very good at predicting the best next word, like you were speaking to a human.
And have you noticed that when an LLM like ChatGPT is providing a response it is giving you one word at a time? Well what it’s doing is constantly applying probabilities to determine each next most probable word in the sequence based on the probabilities of the combinations of words and phrases it has stored for the knowledge domain you are exploring.
With that background in mind, what follows are tips and suggestions to help you build your expertise in working with these LLMs. It will take a bit of time, but you’ll gain in understanding of how to work with them the more you use them. And it will prepare you for working with more sophisticated models that are coming.
For example, you may have heard of AI Agents. These are supposed to be systems that are able to do complex tasks or tasks with many steps and decision points. AI Agents are composed of multiple LLMs structured and combined in a particular way. So, if you learn to work with LLMs you’ll better understand how agents work. And AGENTS ARE COMING and are already here in initial form.
LEVEL 1A: Starting with AI tools such as ChatGPT.
Steps to follow. If you haven’t used LLMs much before:
In a browser window, enter only one of these and start with the free model: “chatgpt.com”, “gemini.google.com”, “claude.ai”, “copilot.microsoft.com”, “perplexity.ai”
Enter your question or response. Keep it simple. Ask one focused question per request or the results may be garbled or unclear.
CAUTION: Anonymize any data you load up, so your organization and personal information are not identifiable.
Provide context before the prompt to get more tailored answers.
Before the prompts above, put in context such as your role, and what you expect as an outcome. E.g. “In the role of a manager for a private farm…” and then the prompt, and
Then optionally, after add: “… provide the result in a graphical image”, or “… explain your reasoning”, or “provide your sources.”
Be watchful of the result because AI models can make things up. This is called “hallucination”.
Keep trying different questions/prompts, keeping the request simple until you get used to using one of them. When you’re comfortable with one to start with, trying putting the exact same request into each one of them and see how the responses are different.