I think the world is divided into two types of people: those obsessed with what creative AI means for their work & future and those who haven’t really tried creative AI yet. To be clear, a lot of people in the second category have technically tried AI systems and thought they were amusing, but not useful. It is easy to be deceived, because we naturally tend try out AI in a way that highlights their weaknesses, not their strengths.
My goal in this post is to give you four experiments you can do, in less than 10 minutes each, with the free ChatGPT, in order to understand why you should care about it. In fact, there are three levels to this AI revelation, and I want to explain why people get stuck at the first two, before giving you activities that might get you to the third:
This is a toy
This is a great way to create lots of written work (at this phase, people start to think about cheating on essays)
This changes so many things
People think it is a toy because the first thing you try to do with AI is what it is worst at: First, you use it like Google: tell me what the best city is, who is the best soccer player in history, and so on. These answers are terrible. They are limited to before 2021, they are often wrong, and they are really boring to read. Right now, AI is not Google. So people leave disappointed.
Second, they may try something speculative, using it like Alexa, and asking a question, often about the AI itself. Will AI take my job? What do you like to eat? These answers are also terrible. Creative AI systems have no personality or sense of self, are not programmed to be fun like Alexa, and are not an oracle for the future.
If people still stick around, they start to ask more interesting questions, either for fun or based on half-remembered college essay prompts: Write an article on why ducks are the best bird. Why is Catcher in the Rye a good novel? These are better. As a result, people see blocks of text on a topic they don’t care about very much, but it certainly looks very much like a good essay or article. That is pretty amazing! But not that useful. They usually quit around now, convinced that everyone is going to use this to cheat at school, but not much else.
But they have only reached the second level of revelation. I want to suggest ways to get all the way to level three.
To be clear, I am not a AI expert, and only understand the details of AI system at a high level. You are almost certaintly not a computer scientist, either. That is fine, because I think some experts are also blinded to what last week’s increases in GPT’s capacity means.
They tend to be worried (correctly!) about the fact that AI is very bad at tasks that involve giving you accurate information you cannot check. An AI will confidently tell you false information in ways that are utterly convincing. That is why it is a terrible Google replacement (or doctor). But most creative and analytical work involves doing lots of stuff where small amounts of errors are common, easily checked, and correctable. Accuracy is not as large a problem as many might expect.
So lets get to the good stuff. Here are four exercises to help you understand the potential impact of AI. For all of them you will need to be signed up for the free ChatGPT system (it is free to sign up for). Lets go…
Path 1: The Invisible, All-Knowing Intern
If you are interested in going into detail on how ChatGPT can help you write, I have a whole post on that.
The first path is to treat the AI like a magic intern with a tendency to lie, but a huge desire to make you happy. To try this path, ask the AI to do some sort of writing you would need for your job. It could be an essay, a memo, or even an email to your team. One powerful approach is to give it a series of bullet points you want it to write:
I tried write an article about why entrepreneurship is hard. Feel free to experiment with your own choices, and some prompts may not work, so you can try variations or use the “try again” button. Get something completely fine, if not inspired. The point is to give yourself a okay starting place.
Now, ask for changes. Remove the paragraph about uncertainty, add one about how hard it is to find an idea. Add a real example case to the point about competition.
You may have to experiment to find the right prompts, but you can correct the AI when it is wrong:
I told you to remove the second point, you removed the wrong one
That isn’t a real company, I told you to use a real company
And then you can start to try all sorts of things:
Now argue the opposite side
Write it in the style of a horror novel. Use more adjectives. Make it vivid. End with a twist.
Make the language punchier and more descriptive
Add a fictional example that explains teh points
Try a different example
Use more active verbs
Turn this into a rhyming poem
Some results will be terrible, or wrong, or both. But it doesn’t matter. You can just ask it to change those things. This interactivity is unlike anything you would get with a real-world assistant, and, with a little knowledge and experience, you can achieve amazing results, multiplying your productivity.
You will also notice that the AI makes mistakes and errors with complete confidence. That is why it is so important for you to have knowledge of the area you are in. You can tell the AI to correct the mistakes, but you need to know enough to understand it. To be clear, creative AI will not make a non-expert an expert (at least in the current version of the system).
Path 2: You are a Game Designer, Now
I just got a text message from my friend Mike. It said, and I quote: “ 🤯 I made a game.”
Mike is a baseball fan, and he asked ChatGPT this: Set up a specific in-game baseball situation and give me four choices that I can make as a team manager.
He had just created a game.
Now what? Well, you can answer the question by picking a prompt, and the AI will continue the story (I would suggest writing: I pick option 4, explain to me, in detail, what happens next. Stop at the next decision, and give me four choices). You can correct mistakes or add material (Add another choice to ask about the option to steal home), you can add detail (Describe who is at bat, Give me the stats on the pitcher) or you can ask it do something else (I bring in the Flash to run the bases faster).
And it isn’t just baseball, obviously. You can ask it to simulate a business situation (I am the CEO of a large company that makes shoes. I hear that real AI has just been invented. I have to decide on a strategy, give me choices on how to respond) or a fantasy game (Set up a fantasy city full of adventure and danger. Describe it in detail and give the choices I can make as an apprentice wizard).
Addendum: Ian Bicking put together a great prompt to get the AI to make Choose Your Own Adventures. Copy it and paste into ChatGPT for an easy way to get a game started.
You can tell the AI to create additional rules, decide actions randomly, or any other option you like. The power of human plus AI worldbuilding is immense.
Path 3: Launch a Business
If this path interests you, I have a whole post on how you can use ChatGPT to help you in entrepreneurship
Idea generation is a place where wrong bad ideas are easy to reject, and may help you think of good ones. The fact that AI can easily come up with ideas, even though some will be bad, makes this another way experience its power. So, lets explore some ideas, with the eye towards launching a business.
Pick an industry you care about, for example if I was a doctor who wanted to stop practicing: Give me 25 ideas about how to make money with a medical degree in an industry other than medicine
Now explore and expand on the things that interest you: give me more information on 15, what does that involve?
That sounds interesting. But how do I explore it? Ask: What would be some steps I can take to test whether this is a good business idea. I have no business experience.
As an entrepreneurship professor, I can say these things are generally good advice. We can go deeper still:
Can you give me a sample business plan?
What is a list of questions I should ask during the customer interviews?
Write me a letter to send to a local hospital pitching my idea and asking if they want to do a small-scale test.
Path 4: Hallucinate Together
One problem with AI is its tendency to hallucinate, making up fake facts that seem entirely justified. This is a problem if you are not an expert in the field the AI is addressing, but it also means that the AI is capable of tremendous world-building. You can work together to go as deep into a world as you would like.
Start with something fun. Taylor Swift starts a supergroup with the Beatles who have time-travelled from 1969. Give me the set list from their first concert, telling me which songs are by which people, and which are new songs. (For Taylor Swift fans, remember that Midnight was released after the cut-off date for the AI’s knowledge. For Beatles fans, I misspelled the band name in my original prompt, and the AI used my misspelling)
You can always tell it to redo the story differently (it originally did not tell me the names of the new songs, so I had to ask redo the song list, filling in the names of the new songs.) We can now start to go deeper:
Write a one-act play about the Beatles first meeting Taylor Swift
What did Stephen Colbert’s monologue say on the day of the concert?
Tell me what happens next
Give me the review of the first concert. From the perspective of Rolling Stone Magazine. From the perspective of a Taylor Swift fan who doesn’t like the Beatles. From the perspective of a mouse who has become sentient and now works for the New York Times.
Other Possible Paths
Automate your job. A recent Substack of mine talked about automating being a professor, you can spend some time seeing how many of your job tasks you can replace.
Learn with it: ask it to explain a concept like you are five, or with clear examples, or with an experiment you can run. Ask it to explain why something is important, or how two concepts are related. Ask it for a syllabus on a topic, or a lecture about it. Ask it to correct a mistake in your code, or your writing. Don’t try this on concepts that you don’t know at all, but it can help you take a basic understanding and make it better.
I really urge you to try one of these paths, or to share others that might have worked for you in the comments. I think understanding the potential of something as new and odd as AI, and especially ChatGPT, is challenging. And this is just the beginning, rumors abound of GPT-4, which will likely be another huge increase in AI ability. If you hit a wall in your explorations now, it likely won’t exist sometime in the next year.
And that is a revelatory realization.
Your 10 minute bet paid out – although in fairness I've already been thinking about this a lot this last week!
I particularly like your ideas on path 3 and 4. ChatGPT in its current state can be very useful as a tool for designing thinking and brainstorming. The point here is to hallucinate at least at the beginning! Using AI as a tool to foster a divergent thinking processes is a great application where the quantity is more important than wrong answers to start. All it takes is one idea that can break you out of a local minimum where you are stuck and inspire a new train of thought.
This is excellent. One thing I wonder about is that the level of this querying process of co-creation requires a certain mindset to start with. Those who would do this seem like they would've done something similar elsewhere (using Google and Wikipedia perhaps) and now use better tools. This seems a smaller group than I'd like, especially re the love of whimsy. I hope this expands thuogh.