Prompt Engineering with Tree of Thoughts

I LOVE asking ChatGPT one question in a variety of ways to see how differently I can get it to answer. This does a few things:

• It forces me to think about what I want to solve for (the problem) up front. So, initially I have accuracy and efficiency in mind, but after experimenting, that always changes because:
• It helps me see answers from different angles that I maybe never would have considered, which sometimes prompts (see what I did there?) me to rethink the question altogether.
• It helps me understand how the AI is interpreting what I’m asking and how I can clarify my question or ask another one entirely.

This got me thinking that I probably should understand if these “experiments” I’m doing have official names or theories behind them.

Spoiler alert…of course they do!

I’ve been watching and reading a lot about prompting techniques, and it’s awesome to be able to describe these things I’m doing with concrete names, even if I’m not quite executing (or, honestly, understanding) them perfectly. I’m learning that I can get direct, straight-to-the-point answers at first, but then I can get it to give me diverse points of view. And let’s face it — in this day in age, getting multiple viewpoints on a particular topic is important. AI can help with that, particularly if you’re prone to getting stuck in your own stubborn brain.

::Whose brain is stubborn? Mine? Noooooo.::

I learned about zero-shot, few-shot and chain-of-thought techniques, but the one that really intrigued me was the tree-of-thoughts prompting technique.

I began by asking the classic ToT prompt:

Imagine three different experts are answering this question.

All experts will write down 1 step of their thinking,then share it with the group.

Then all experts will go on to the next step, etc.

If any expert realizes they're wrong at any point then they leave.

The question is…

My question was: How is global warming affecting Earth?

The answer gave me some interesting insights from the viewpoints of three different experts: a climate scientist, and ecologist, and an economist.

It was awesome to see the different viewpoints from different “experts,” and while there was no concrete, black-and-white answer, per se, it was interesting to see the collaboration show answers from scientific, ecological, and economic dimensions.

I went a step further and responded with, “Now, reimagine this scenario with a democrat, republican and libertarian.”

This was very helpful because I’m this classic persona: “Let’s visit the comments section of a partisan post on social media to see how everyone hates everyone else and has an opinion and no one can talk about this in a civil, respectful manner.” It was refreshing to read different viewpoints without anger and vitriol behind it. And, while I might not agree with how each “expert” responded to the prompt, I still learned diverse viewpoints without the worry of stumbling across keyboard warriors wanting to argue without calm logic.

I’m going to keep playing around with this and researching to see how else I can use this type of prompt to help understand how customers might think in different ways. It doesn’t even have to be correct answers — I think if it’s helping me learn to think outside of my own brain box, that’s a win.

My prompting journey continues. If you’re stumbling across this blog from who knows where, please be patient. I’m learning! I’m open to feedback, always! Just click “contact” in the menu above. I play around mostly on ChatGPT because I have a paid subscription.

Inspiration:

Advanced Prompt Engineering Techniques

Tree of Thoughts Prompting

Prompt Engineering Guide (ToT)

The Revolutionary Approach of Tree-of-Thought Prompting in AI