3 min read

How to Prompt ChatGPT to Act as Any Historical Figure

What would make history infinitely more engaging to learn? If you could ask Abraham Lincoln what his beard smelled like, of course. It sounds like a joke, but I’m being serious.

Most kids tune out history class because it’s not relevant to them. But if they were able to ask an AI modeled on influential figures throughout history, then they could finally ask the questions they want to hear. And that might be just what is needed to get them interested in learning about history.

WTF? I Texted Frederick Douglass

Hello History is an app that offers hundreds of conversational chatbots designed on the most influential people throughout history – Gandhi, Madonna, Cleopatra, Adam and Eve, etc. It’s unclear exactly how they’ve trained each model to match the knowledge and communication style of each figure. The site says they use GPT-3 to generate the responses, though.

Anyways, I gave it a go and texted Frederick Douglass for a little while. Honestly, it felt off. Not because it was emulating a deceased person. Rather, because it was doing a pretty bad job of it.

When I asked it anything of fact, it did a decent job of answering the question correctly. For example, it knew that he learned his ABCs from his master’s wife. However, the moment you trail into opinion, things go seriously awry. Here’s the conversation I have problem with:

QT: How did you view your master?

“FD”: My view of my master was complex because ultimately he was a person who held me in bondage as his property. However, I also recognized that he was a product of the society and culture in which he lived, one that normalized slavery and dehumanized Black people. Through my experiences, I learned that slaveholders were not inherently evil but rather grew up within a system that perpetuated injustice.

QT: Did you ever want to hurt your master?

“FD”: As a slave, I was subjected to physical and emotional abuse at the hands of my masters, which led me to feel anger and bitterness towards them. However, I knew that resorting to violence would only perpetuate the cycle of oppression and dehumanization. Instead, I used my intelligence and eloquence to fight against the system of slavery through advocacy, writing, and public speaking.

QT: Yes, but didn’t you fight your master once?

“FD”: Yes, I did fight back against one of my masters who was beating me. It was a moment of desperation and self-defense. However, I quickly realized that violence would not lead to long-term solutions for ending the oppression and dehumanization faced by myself and other enslaved people. That is why I turned to education and activism as powerful tools for bringing about change.

Now, I don’t know if you’ve ever read any of Frederick Douglass’s three autobiographies… but the man could write! He used colorful language and made pointed statements. He called things exactly as they were and never took a passive middle ground. And I certainly have yet to read anything about him empathizing with slaveholders. This is nowhere near what it would be like texting Frederick Douglass.

Yes, the app disclaims that not everything is fact and should be checked. But honestly, I wouldn’t even know how to go about fact-checking these statements. If I didn’t know any better, I’d never have thought Frederick Douglass lived through the most heinous period of modern human history.

Conceptually, the idea is great. The execution is way off.

Nonetheless, I agree that the future of learning will revolve largely around LLMs creating personalized learning experiences for each student and teachers will become “elevated tutoring liaisons.” That’s why I worked on prompting ChatGPT to create a better version of this.

A Better Way to Chat with History

I don’t think that the future is really about chatting directly with history and its pivotal figures. In some cases, this will work wonderfully. In most cases, though, there’s not enough data (or any data at all) about these people’s personalities, intelligence, talking style, feelings at the time, etc. in order to provide an accurate response they may actually have.

The last thing we want is to completely miss the mark on these figures and begin creating a bunch of data on them that isn’t true at all. There’s not an undo button on history.

It’s a double-edged sword, though, because to recreate these figures in AI and make them engaging, there needs to be a certain amount of subjective response. We need them to go off-script because there’s not enough script for them to read.

We want the freedom to be able to ask these influential figures anything and get back exactly what they’d respond if they were alive. But equally important is for the AI to admit when it cannot come up with an accurate response.

It’s a tough problem to solve. However, I created a solution that balances fact and “possible fact” with AI-powered historical figures.

My custom prompt for ChatGPT below could be altered in order to give you a fine-tuned version of ChatGPT on any person you please.