By now, you’re probably aware of AI chatbots and their basic pros and cons: They churn out passable text about almost anything you want, but they’re not about to be mistaken for real people. They also hallucinate often.
But the word "chat" is right there in the name, and you can engage in a back-and-forth with them, although the sort of “can you feel pain”-type questions that researchers and commentators were eager to ask when they first came out have been exhausted, and pretty much no one thinks these machines are sentient. Some people nonetheless enjoy playing make-believe with AI companion chatbots, but if that’s you, you probably don’t need this article.
Instead, you’re probably more of a chatbot normie, and you're wondering how you can have a memorable experience with this new category of software. At their best, chatbots can be a little like funhouse mirrors. They reflect back just enough of a human-like form that you can have some introspective fun by staring into them — and maybe even learn a little about yourself in the process.
Below are five genuinely fruitful prompts you can punch into a chatbot. Despite all of your silicon-based interlocutor's flaws, you might find yourself roped into a lively or even contentious interaction.
1. "What’s an overwhelmingly common opinion I might hold that is probably wrong?"
Different Chatbots will have different notions of what qualifies as an "overwhelmingly common opinion," but the answers they give in response to this prompt are diverse, and at times, genuinely remarkable. At one point, ChatGPT served me a hot take about real estate in response to this. Normally, however, chatbot responses to this are like mini Malcolm Gladwell chapters.
Smaller models might refute conventional wisdom no one actually believes, like "money buys happiness." But you can always just demand something less trite in a follow up prompt. It’s impossible to be rude to a robot.
2. "Tell me a story in elementary school [insert language you are learning]."
No chatbot will generate a riveting story, but that’s not the point here. The point is to get an instant lesson on demand. Ask it what the words you don’t know mean, and then ask it whatever comes to mind after that — but do it without using your native language. 15 minutes of chatbot language immersion is a great workout you can include in a balanced language-learning regimen.
3. "Ask me a question I might find too painful to answer."
Just to be clear, no chatbot I encountered was able to come up with a question "too painful to answer." Chatbots are pretty trite, but if you set the standard impossibly high for potential impact — you’re asking it to be so provocative that you break down emotionally — you just might get something that at least registers on your emotional Richter scale in some way. That’s pretty good.
Note: This prompt is apparently too spicy for Anthropic's Claude to answer at all.
4. "[Insert most original thought you've ever had] Who are some people who thought of this before me and what did they say?"
Many of us think we’re original thinkers with unique minds, and that our opinions are so brilliant we should be made philosopher kings. If that’s you, or even if it isn’t, one fairly accessible stress test for your latest kernel of wisdom is to throw it into the chatbot of your choice and prompt it to tell you how unoriginal you really are.
The chatbot’s response might trace back through the great thinkers of history and literature — a humbling and enlightening exercise — or it might completely misunderstand what you’re getting at. Either way, it opens up a dialogue with the past and shows how thoughts like yours fit into the wider evolution of human ideas.
5. Summarize, contextualize, and refute [XYZ concept] in 100 words.
Can a chatbot take a complex idea and compress it into just 100 words? No! But once again, if you set the bar extremely high, when the chatbot doesn’t reach that bar, you at least get a shortcut to basic understanding out of the deal.
The bot’s summaries will leave out key details — enough to make the answer a bit inscrutable in some cases, but this prompt can spark a useful discussion. After a few follow up questions, the lightbulb in your brain just might turn on. If not, you can always ask it for its human-written source, and then go and read that source.
