5 Lessons From Experiments With ChatGPT

I’ve spent the last year trying to navigate the Matrix of generative AI.

About 7 months back, I decided to delve deeper into the poster boy for the Gen AI scene in ChatGPT from OpenAI. These are the 5 lessons learnt so far in this journey of self experimentation and guidance from smarter people than I.

1/ It’s Narrow AI

✍️ ChatGPT is a ‘Narrow AI’. That basically means it focuses on one task/s only. It’s limited and needs supervised learning from you (the human).

🤖 Artificial General Intelligence (AGI) is what you see in the movies like Terminator and Ex-Machina. This is a self-aware form of technology. One that thinks and learns independently. It does not need human input.

Generative AI tools such as ChatGPT will change the way we work and even search for information. They won’t be taking over our world anytime soon (at least I hope so).

2/ Never Prompt and Ghost

ChatGPT is not a mind reader.

I’m sorry to break that to you. Nor is it some kind of magic where it just ‘gets it’. If it helps, imagine CGPT like an intern.

They don’t know what they don’t know but you can help them unlock their capabilities with the right instructions. You wouldn’t talk to them once and then give up if they didn’t get it the first time, right?

The secret sauce is in building an ongoing dialogue between you and the tool.

3/ It Needs Clear Input

Better input = better output.

If you were tasked to complete a big project you had never done before and were only given one sentence of information, how would you feel? Lost? Exactly. Now you know how CGPT feels.

Prompts need 3 things to do well:

1️⃣ Specificity

2️⃣ Context

3️⃣ Constraints

4/ It Needs A Human Touch

60% = CGPT, 40% = Human.

ChatGPT is not a silver bullet. It’s not going to create high-quality content as good as a human. It doesn’t think critically and has no clue about real emotions.

But you do.

You must co-create not delegate. More on that in our fifth and final lesson.

5/ Review, Edit and Challenge

Never use any output from ChatGPT without:

• Reviewing content
• Editing for your audience and style
• Identifying and challenging bias
• Fact-checking research and data

Large language models like chatGPT are only as good as the data they’re trained on. So, if that’s rubbish then you must be cautious.

Generative AI can and will hallucinate. It’s up to you (the human) to always review, edit and challenge outputs.

Before you go… 👋

If you like my writing and think “Hey, I’d like to hear more of what this guy has to say” then you’re in luck.

You can join me every Tuesday morning for more tools, templates and insights for the modern L&D pro in my weekly newsletter.

You might also like

The ChatGPT Crash Course For L&D Pros: Future proof AI skills in hours, not days.

Generative AI Explained: A simple guide for humans

The Case For Relevant and Accelerated AI Education